Marvell Technology, Inc. (MRVL) Earnings Call Transcript & Summary

Good morning, and welcome to Marvell's 2021 Investor Day. Let me give you a quick preview of today's presentations. Matt will discuss our plan to capitalize on the cloud-driven disruption in the semiconductor market. Raghib will lay out our technology platform to deliver silicon optimized for cloud customers. Note, we'll take a short break after Raghib's presentation. Loi will then take a deep dive into our market-leading electro-optics platform. Dan, Nariman and Raghib will discuss the product leadership and share gains they are driving in our storage, networking and compute businesses, respectively. Note, we'll take another short break after Nariman's presentation. Jean will bring it home with our plan to deliver strong long-term financial returns for our shareholders. We will wrap up our presentation with a Q&A session. [Operator Instructions] Before we get to the fun stuff, let's go over our safe harbor statement. Except for statements of historical fact, this presentation contains forward-looking statements but -- including, but not limited to, statements related to market trends and to the company's business and operations, business opportunities, growth strategy and expectations, and financial targets and plans that involve risks and uncertainties. These statements are not guarantees of results and should not be considered as an indication of future activity or future performance. Actual events or results may differ materially from those described in this presentation due to a number of risks and uncertainties. For factors that could cause Marvell's results to vary from expectations, please see the risk factors identified in Marvell's quarterly report on Form 10-Q for the fiscal quarter ended July 31, 2021. During this presentation, we will be making -- mentioning certain non-GAAP financial measures. A reconciliation to GAAP is provided at the end of this presentation, which will be posted to our website. Let me now explain the syntax behind our fiscal calendar. Marvell financial measures in this presentation correspond to our fiscal calendar. Our fiscal year generally starts in February and runs through the end of January of the following year. As an example, fiscal year 2021 started in February of calendar 2020. Let's kick things off now with a short video, followed by Matt's presentation. Thank you. [Presentation]

Good morning, everyone, and welcome to the 2021 Marvell Investor Day. If you look at the world's most valuable companies, 7 of the top 10 have a combined market cap of over $10 trillion. They all have 1 thing in common, the cloud. The cloud is the single most disruptive force in the global economy over the last 10 to 15 years. And the single most disruptive force inside the cloud is the silicon on which this data infrastructure is built. It's called cloud optimized silicon, and it's already happening. We're going to talk to you today about how Marvell is perfectly positioned to capitalize on this disruption and massive growth opportunity with a portfolio that spans compute, networking, storage, security and electro-optics, and a customer engagement model that is unique to Marvell. If you want to invest in the future of the cloud, Marvell is your company. Just to give you some context, when I talk about the cloud, I'm really talking about the entire data infrastructure market. There are 4 markets that Marvell serves. You're going to hear about this today from our team and every one of these has a cloud strategy behind it. In the data center, this transition is obvious. Carrier infrastructure is going to virtual RAN and cloudification. You'll hear from Raghib about this today. Automotive with 5G and autonomous driving, all need a cloud strategy and the enterprise has become borderless. The bottom line is all of these end markets are converging to the cloud. So let's see what's driving the cloud silicon market. There was a belief going back 10 years ago, maybe more, that the cloud would run on industry standard hardware powered by x86. Just think about it as taking existing on-prem data center hardware and scaling it up to the cloud. Software was supposed to do everything, turned out to be a very inefficient solution when you think about the scale required in cloud computing. What you hear about today is something called heterogeneous computing, a concept where you have specialized processors for different types of workloads and applications. GPUs for parallel workloads such as machine learning, FPGAs for acceleration. Each of these solutions are very good on their own merit. But if you look at the problems the cloud customers are trying to solve, they need even more. What I'm here to tell you today is that the opportunity is cloud-optimized silicon, silicon that is highly customized for individual cloud customers and applications. The companies that enable this are going to be big winners. And it's not just happening in the compute. It's happening in the storage. It's happening in the networking. It's happening in the security. And it's happening in electro-optics. So to really address the entire opportunity, you need to be excellent at each one of these. I had a realization several years ago when Raghib and I went to a customer meeting together, and we were driving back in the car, just the 2 of us. And we had both reached the same conclusion in the meeting. He turned to me and said, "Matt, I think the future of this market is going to require customization and optimization, and we need to prepare for that." And I had reached the same conclusion in the meeting, which was the notion that selling a merchant or general purpose product wasn't where the market was going to go. It was interesting because just a few miles away at Inphi, Ford Tamer and Loi and Nariman and his team all reached the exact same conclusion, and that's really what led Inphi to their incredible growth. It was their recognition that this market was moving to an optimized model. Let me be clear, this is not a zero-sum game. There will be -- continue to be a very healthy market for general purpose products in the cloud. This is about enabling new capabilities in the cloud. And where -- and heading to where the puck is going. So let me explain why this is happening. It's because every cloud is unique. The architectures of these large cloud companies are all completely different. They actually design and build their individual data centers optimized for their own specific applications. Let me give you 2 examples. If you're one of the cloud titans and your leading franchises search like Google, you have a very different architecture than a company like Facebook that's serving up photos, videos or advertising. They optimize their cloud data center design for the applications that they serve. One thing is clear, each of them have their own unique requirements and massive scale. So let me give you an example of cloud optimized silicon, what that means. On the left, you see a representation of a general-purpose product. Each of these blocks represents an individual functional block inside an SoC, for example, an ARM core. This type of part represents a superset of all of the capabilities you might have in a chip to solve all of the problems for every data center it's sold into. It's designed for everything and optimized for nobody. This works well in the enterprise market, by the way. But when you get to the scale of the cloud market, there's an enormous benefit in creating an optimized solution. Cloud companies want purpose-built silicon with exactly the right amount and type of each capability. For example, one might need a special-purpose machine learning block built in. Another might need proprietary security functions as an example. It all depends on the use case. So when we build cloud-optimized silicon, it's optimized for each individual cloud, each workload, each application. It's not just all about the silicon though. It's about the software, the firmware, the platform, the form factor in which the total solution is delivered. These are all critical aspects to actually optimizing and customizing these solutions. At Marvell, we have cases where we're selling a card with multiple Marvell chips with all the software, all the firmware, everything tuned and optimized to the customer's application. That's an example of cloud optimized. And don't just take my word for it. You can read about this in the press on a regular basis, sometimes a daily basis. It's clear these cloud companies are very active in silicon. Now some of you may be wondering, okay, is this a good thing? Is this a bad thing for Marvell? It's completely aligned to our strategy for cloud optimized silicon. And this is good for Marvell. As you will hear throughout the presentations today, Marvell has all the pieces to be the indispensable partner to these companies to allow them to achieve their ambitions in silicon. Now let me show you how enormous this opportunity is. Last year, we told you the TAM for infrastructure was $110 billion, growing at 6% a year. We said the Marvell portion was a $16 billion opportunity growing to about $20 billion. which was growing faster than the TAM. Okay. This year, I'm here to tell you that the infrastructure TAM has grown close to $130 billion. And as a result of the market continuing to accelerate, more exciting is that Marvell SAM is now 50% larger than it was just 12 months ago when we had this Investor Day. It's driven by our SAM growing into the overall TAM on an accelerated basis, and the transformational acquisition of Inphi and the recent addition as of yesterday of Innovium. And the growth rate is also accelerated for Marvell, from 9% a year compounded annual growth to 13%, giving us a $30 billion opportunity in just a few years. And let me put this in perspective. We had our first Investor Day in 2017. We identified at that time, the total market opportunity for Marvell was $8 billion. The market is now $30 billion that we're going after. It's almost a factor of 4 increase, larger than just a few years ago. So we're in an end market that's growing 13% a year compounded off a huge $20 billion base. We've taken a look at all the major industry peers, by the way. We can't find a company that's got this kind of end market growth at this level. As you can see, the data center is the largest of our end markets. It's growing 24% a year off a very large base. Within that, cloud is actually growing faster. Within carrier, we're focused on the 5G opportunity. And as you can see in automotive, it's a smaller market, but it's the fastest growing. So if you put the company's growth drivers, which are cloud, 5G and automotive together, those end markets are growing 20% a year, even faster than the 13% overall SAM. These markets represent the largest opportunity for Marvell. Now let's take a look at how we've transformed the company to address this very unique opportunity. 5 years ago, Marvell was a consumer company. Today, that's completely inverted. We've pivoted a data infrastructure that we've been investing to become the leader. And now the vast majority of our revenue is in infrastructure. To be clear, the market didn't just move Marvell here. What you see here is the result of a very deliberate plan over the last 5 years and strong execution to get to this point. Now let's see how we compare with our peers. So let's take the revenue split of infrastructure versus consumer, 84%, 16%. And let's compare it to our large industry peers. First observation, Marvell has, by far, the most exposure to infrastructure and the least exposure to the consumer market. This is important because the infrastructure market has extremely long product life cycles, sticky designs, high level of visibility and outsized growth. By contrast, the consumer market by nature, is more volatile, more short term. You need to win new designs every single year. So we chose a different path. We also have the largest exposure of any major semiconductor company to data center, which as I showed you earlier, is the fastest-growing portion of the entire market. It's the most lucrative, disruptive end market opportunity in the industry. So if you believe data infrastructure is the most important growth driver in semiconductors over the next 10 years, Marvell is your company. Here's how we deliver on it. First, we have the leading technology platform bar none in the industry for infrastructure. What you're looking at here are all the critical building blocks you need for this market. And the way we got here was not by accident. We've been very deliberate in our approach to build this company to service this opportunity. So we started with 3 key areas of organic investment. We doubled down on storage 5 years ago. You're going to hear about Dan from that. This was a business that was in decline, and it's now growing double digits. We doubled down on enterprise networking. And you can see from our financial results, and we actually break this out that our business is outgrowing the market. And third, we had -- 5 years ago, we had 0 revenue in automotive, Ethernet. And now it's a booming business, and you'll hear from Nariman today about that opportunity. Next, we did 2 transformational acquisitions to drive our growth. The first was Cavium in 2018 and the second was earlier this year with Inphi. This gave us the compute capability and the electrooptics capability we needed. These were complemented by 3 very other strategic tuck-in acquisitions of Aquantia, Avera and Innovium. We also decided what not to do, and we executed multiple divestitures that ultimately generated $2 billion in cash for shareholders, and it helped hone our focus. The pieces of this platform, they existed, they were just inside of their companies. They were inside Marvell, they were inside Cavium, they were inside Inphi. We brought all these teams, the assets, the technologies, altogether under one roof to execute on this opportunity and the whole is much, much greater than the sum of the parts. We worked incredibly hard over the last 5 years to build this company, and now we're looking forward to reaping the rewards and scaling the company for growth. We have all the pieces to win. Here's how we scale. It starts with the foundation of leadership products and technologies, how we work with customers and enable them to succeed, scaling up our operations, putting in place the right capacity to grow the company. And finally, none of this can happen without a winning team and culture. We're investing in each of these critical areas. So let's take a look at each one. It all starts with the foundational process technology, platform and IP. Last year, we told you that Marvell was going to be a leader, not a follower in process technology. Our 5-nanometer platform is off to an incredible start. We already have over 20 programs in development with customers. Last year, we talked to you about test chips. This year, we have product silicon back in the lab on multiple products with an intent to ramp those into production next year. And you can bet that we're planning to be a leader in 3-nanometer as well and keep driving our technology platform forward. Product leadership is much more than just technology leadership. Product leadership is winning product definitions, winning features, understanding the customer requirements and putting it into the products. When I look across our entire portfolio now, Marvell has leadership products in every single category where we compete. The output of all this is what we call market leadership, which is execution, generation after generation after generation, building customer trust. This is how we win. Customers are at the center of our strategy. We're building true partnerships with leading companies in our markets, not just the big incumbent giants, we also focus on the innovators, the disruptors and we're taking the long view to enable their success. Last year at Investor Day, we told you we had a path to 13 customers that would generate over $100 million a year in revenue, which is a significant milestone when you look at a customer relationship. When we announced the Inphi transaction, we increased that number to 17. We're well on our way. And today, we now see a path to 19 $100 million-plus customers for Marvell. This is all aligned to our flexible business model, which is unique to Marvell. We can enable customers to go from full custom design to full merchant offerings to everything in between. Often, that conversation starts with ASIC or custom, moves to one of the other business groups. We look to see who has the right IP, experience and know-how and the ability to do this is unique for Marvell. These are highly strategic, confidential relationships we have with these key companies. And while we're going to talk to you today about some very exciting projects we're working on, we maintain the highest levels of customer confidentiality, and won't be disclosing any specific customer names. Thank you for understanding that. Each of these cloud optimized programs, they come with significant NRE, where they are partially funding the development of the chip. This expands our R&D footprint and our ability to invest. The reason is because of the value of our IP portfolio, our architecture expertise and our track record of execution. Our customers view us as a true extension of their engineering team. These are critical customer engagements when you get to $100 million relationship or in some cases, a $200 million-plus relationship. The projects that we take on, we execute are typically 5 to 7 years in production. Customers care more and more about ensuring that their long-term supply needs are met, and Marvell is putting significant investment in people, tools, technology and also long-term agreements with our supply chain to ensure our and, more importantly, our customers' growth plans. I can tell you that Marvell is viewed by the supply chain as one of the most attractive companies in the industry to work with. They like our business model and our long-term focus, and I'm confident that we're going to be able to secure the capacity we need. We've got this. It all starts with the team. The talented group of people we have working at Marvell. And we have the team we want. I'm blessed. We have an incredible group of leaders in place, with the best business and technical capabilities anywhere in the industry. And it's not just my direct team, that strength extends to the next 100 key leaders and even more broadly, across the company. Marvell is a company that people want to work for. We've become a destination for talent in the industry. And while we've come together from many backgrounds, we have a strong company culture that binds us together as a team. Our core behaviors are the foundation of that culture. It's all about treating people with respect and creating opportunities, including those with diverse or underrepresented backgrounds. This has been a key competitive differentiator for Marvell. And on that point, we've made substantial progress on our inclusion and diversity program. We've hired a great I&D leader, and we've launched our first global strategy. We're focused on activating and empowering leaders, creating an inclusive place to work and cultivating a diverse workforce. We're also very committed to a set of initiatives focused on improving opportunities for women in the company. In the past year, we've expanded online reporting of gender and diversity data. This is hugely important to our stakeholders, including our employees, our customers, our shareholders. It's hugely important to me personally, and we're making this a priority at Marvell. If I use a sports analogy, if you want the best team, you need the best athletes. The companies that can attract and retain and enable the best talent in the world will win over the long term. And one of the things that matters most in my experience in attracting and retaining talent is that people want to be on a winning team that have the best products, the best technology, market leadership. That said, you also need to be a great place to work and treat your employees well. And don't take my word for it alone. We've been recognized by our industry peers. The GSA last year named us as the most respected company in the industry. We've been recognized by our own team as the best place to work. Our employee SAT and Glassdoor rankings and ratings continue to rise. And we've been recognized by our community for philanthropy, and even by the financial community with multiple categories on the institutional investor All-America list. We've been very focused on building the Marvell company culture the last 5 years, and it's great to receive these external recognitions. Climate change is one of the most pressing issues facing the world. And as a large and important company, we have an obligation to do our part. Today, I'm here to tell you that we are committed to net zero emissions as a company. We're adopting a science-based approach to do this. And over the next year, we'll lay out our time line for achieving these targets. And like everything we do at Marvell, we're going to measure, track and improve and report out the progress we're making. I consider this to be a business imperative. It's [Audio Gap] disruption. We have a winning strategy and all the pieces in place. We're scaling to the growth opportunity in front of us. That [Audio Gap] growth and value. So let's talk about how this translates financially. [Audio Gap] the financial community on over the last few quarters. It's going very well, and our team is making great progress. Some of you might recall that we said by the time we reach fiscal '25, fiscal '26 time frame, the incremental revenue from these wins would be roughly equal to Marvell's entire cloud revenue last year, which was about $400 million. Today, I'm here to tell you that we're on track to deliver, but now we expect to deliver on it. More exciting still is that based on new wins and higher demand for our products, we expect that the time we reach fiscal '25 and '26, the incremental revenue that we've won will actually double in that time frame. This is very exciting. So let's put this in the context now of the overall market opportunity. Again, we have $20 billion opportunity. It's growing 13% a year. We're investing to drive outsized returns in cloud, 5G and automotive, and those markets are growing at 20% a year, and we're winning. Today, we have $1.5 billion annual run rate in these markets. So just like big round numbers, cloud is about $1 billion run rate today; 5G, $400 million; and automotive of $100 million. So that gets you to the $1.5 billion. We can -- believe we can grow this market, grow Marvell at 2x the market, which is what we actually said a year ago, although the market growth rate has actually increased. Now this includes the Innovium acquisition, which we just announced yesterday and all those new design wins I told you about. This is huge growth. In the rest of the infrastructure market, we have over $2 billion in revenue today and expect to outgrow the market as we take share in enterprise and in the wired market with Inphi's unique optical products that Loi will talk about later. We have a great opportunity. There is tremendous momentum overall in our business. Last year, we outlined 10% to 15% long-term growth in our financial model. As a result of the market acceleration, the acquisitions of Inphi and Innovium and stronger demand for our products, we are significantly increasing our long-term growth target model to 15% to 20%, which I think is one of the highest in the semiconductor industry. Now I don't want to steal any thunder from Jean's presentation, but you will see that we expect to drive this level of growth with an industry-leading overall financial model, revenue growth, gross margins and operating margins. The future is very bright at Marvell. I hope you enjoy the rest of the day, where you'll get to meet my team, including some of our new members that joined us from Inphi. And I hope that by the end, you will be as excited about the future of Marvell as I am. Thank you very much.

Thank you, Matt. Good morning, everyone. It is great to be with you here today. As most of you know, I came to Marvel through Cavium acquisition, where I was Co-Founder. Recently, I took a broader role at Marvell as President of Products and Technology. As a part of my this new role, I'm responsible to coordinate across all Marvell business units and the customer to make sure that the product that we develop is in full alignment with the needs of end markets and customers, and they work together to deliver the optimized solution. Today, I'm going to talk about the disruption in data infrastructure by the data economy, and how Marvell is very well positioned to take full advantage of it. Matt talked about how every cloud architecture is unique and requires optimized silicon. So let's take a look what is a cloud architecture. It has lots and lots of servers to provide compute, security and storage. These servers are connected together with a switch on the top of the rack. And then there are multiple layers of switches to connect all the racks together. Now these switches are connected together with optical interconnect with optical modules on both sides. And then in a cloud, there are lots of data centers connected through optical interconnect. The unique part about Marvell is that we have market-leading products for each one of these capabilities. We have compute and security, and I'm going to talk more about it later today. We have storage, and Dan is going to talk how Marvell has invested to deliver optimized storage solution for cloud. We recently announced -- we recently closed Innovium. And with that, we have that data center switch. And later on, Loi will talk about how we are leading in electro-optic products for cloud data infrastructure. If a company only has a specialty in one of these areas, and the capabilities of delivering only that product, that is all they see, and that is all they understand. For example, Marvell had storage in networking, Cavium had compute and security, Inphi had optical interconnect and Innovium had switches. We had these market-leading products. But together, we are leading the complete portfolio for the cloud infrastructure, which is second to none. The power of this unique architecture is the vantage point for Marvell. Marvell has all the essential products for every capabilities in the cloud. So at a high level, all cloud architectures look similar. They all need compute, security, electro-optics interconnect, switches and storage. But as Matt talked about, each cloud architecture is unique. And it is unique because it has a unique workload. Let me show you how 2 different applications are driving different needs in cloud architecture. We have search application, which requires a very high performance compute, and it is latency sensitive. On the other hand, we have social media application, which has a lot of requests coming in, but latency is not that critical. At the same time, search works with a lot of data and there's a lot of data movement within the data center. On the other hand, social media has a lot more movement from within the data center to out, not that much within the data center. As a result of that, we need very high-performance CPUs for search application, but for social media, you can have optimized compute but a lot of them. Similarly, we need high bandwidth, low latency for data movement within the data center for search, but social media is not that much important, but you need to connect a lot of servers, you need a lot of ports in the switch. At the storage itself, for social media, you are looking for the capacity for search, you're looking not only capacity but the latency as well. So as I said earlier, at a high level, it seems like all cloud architecture are same. But in reality, they are configured and optimized based on the target workload. Matt shared with you what is needed to implement a cloud-optimized silicon. If you want to be a leader for cloud optimized silicon, you need to be at the cutting edge. You need to be driving process node technology, high bandwidth, high performance, complex package, you need it all, you need to be at the cutting edge. Let's take a look how Marvell is doing here. Marvell used to be fast follower when it comes to process node. But we decided if we want to be the leader in providing optimized silicon for data infrastructure, we need to be at the latest process node. This is why 3 years ago, we made it a company focus and invested to become the leader in process node technology. Today, I'm pleased to share with you that we have established ourselves process node leader in 5-nanometer. We have complete 5-nanometer platform with all the key IPs. We are using this platform to implement our own products as well as offering it to our customer to achieve their own optimized silicon needs. This has been a key part of our technology, and it is driving design win, not only in cloud and 5G markets but also in automotive and enterprise. We are taking several products in production in 5-nanometer. And we are planning to tape out another 20 products in the next 18 months. But it does not stop there. We are fully committed to keep the leadership in process node technology. We are heavily investing to -- in 3-nanometer. You may have recently seen our announcement of partnership with TSMC about 3-nanometer. It is not only about knowhow of 3-nanometer, it is about developing complete platform and the needed IPs to keep the leadership in the process node technology. One of the key critical IP for -- to have the high-performance silicon is SerDes. Just to give you an idea, if you do not have right SerDes, you cannot have a high-performance silicon. This is why the quality of SerDes is important. And that is the reason when we decided to take the leadership in process node technology, we also decided to take leadership on all the critical IPs. And SerDes being one of them, we were the obvious reason to invest in it. So we invested to take the leadership in 5-nanometer 112 gig SerDes. And today, I'm really happy to share with you that we have the industry-leading 5-nanometer, 112 gig SerDes in-house today. We have used this SerDes in multiple of our own silicon, but this has been a driving factor for our design wins with the customers as well. We have won multiple designs. And customers, when they're looking at a SerDes, they look for a few key parameters. They are looking for lower rate, long distance, performance and power. Two of our customers did complete evaluation of all the available 5-nanometer, 112 gig SerDes in the market. And they found that our SerDes is the best-performing SerDes. That is why they chose us, and that is why we won the design. We are not stopping here. Just like we are investing in 3-nanometer process node technology, we are also investing in a 224 gig SerDes. Marvell is fully committed to keep the leadership in process node technology and critical IP. Just like process node and critical IP, package is becoming another critical building block of delivering optimized high-performance silicon. Just to give you an idea, most cloud silicon ends up being multichip on the same package. When you are integrating multiple chips of -- multiple dies on a single package, it is very, very complex. That makes the design as complex as the silicon design itself. There are various examples of multichip packages, for example, 2D, 3D, multichip module, CoWoS as well as chiplet. Package design is becoming a key defining factor to achieve the performance in high-end silicon. We have 25-plus years experience of delivering complex packages and multichip design. We have shipped over 1 million multichip silicon in production. Loi is going to share with you later how optics are getting closer and closer to the silicon. It is called co-packaged optics which means the optics and the silicon are in the same package. Now how do you put the light and the digital logic together in a package? It is very complex, very complex technology. But we have these capabilities because of expertise that we have in-house now combined with the optical expertise from Inphi. Marvell investing in all types of advanced complex packages, and we will be ready to deliver the most optimized high-performance silicon for the cloud data center needs. There are 5 essential capabilities needed for cloud data center architecture. It is not enough to have a process node leadership and best IPs and go to the customer and say, "Hey, what do you want?" That's not going to work. To deliver the optimized silicon for cloud architecture, one have to have in-depth knowledge of technology and expertise in every area of capabilities. And a silicon provider can gain that capability only through implementing market-leading products. We have market-leading products in each one of these areas. We have storage. We have networking. We have compute. We have security. We have electro-optics. So do you know why do we win? We win because we have invested in market-leading products. We have hardened it. We have field proven. And we have deployed in volume production in many, many use cases. And this is why when we go to a cloud provider and present our product offering, at times they use our standard products. But often, they tell us, Mr. Marvel, we like these products, but the problem that we are trying to solve requires the characteristics of the functionality from multiple of these products. And as a result, we believe that it will be great if our architects sit with your architects and come up with the right product design. The discussion has moved from the relationship of customer vendor to co-design. And this is a unique position and advantage that Marvell have. For example, one of our cloud vendor wanted to implement and optimize storage accelerator. They needed expertise in the storage, SSD and searching in security and in ARM compute. And guess what, Marvell has leading products in each one of these areas. And that is why we were the right partner to develop such a product. And it does not stop there. We have established a really great relationship with our cloud customer. We have really become the extension of their engineering team. So now we are having a discussion about what problems that they are facing and how can we work together to bring the best optimized solution for that. Just in the last 6 months, we have been reached -- approached by multiple customers to do a proof-of-concept solution to figure out a solution of some complex problems that they are facing. We are innovating together, and we are developing the most optimized products for their needs. So the cloud is driving the innovation. Marvell is investing in process node technology and IPs and the market-leading products to deliver the most comprehensive product portfolio to serve the needs of the cloud. However, this investment is equally applicable to all of other end markets. The tremendous leverage in R&D that we are getting across all these markets. The same building blocks that we develop for cloud is equally applicable in a data center, in the carrier infrastructure, in automotive and in enterprise. With a strong product portfolio as a foundation, Marvell is driving innovation and product leadership in all of our end markets. Throughout today, you'll hear from my colleagues how we are serving each one of our end markets. Thank you very much. [Break]

Good morning, everyone. I'm Loi, co-founder of Inphi. Electro-optics has been my lifelong passion. I studied at Cornell, got a PhD there and spent the next 12 years working at [indiscernible] as a research scientist, and then the next 20 years at Inphi as Co-Founder. At Inphi, I led the broadband analog and optics business, built it from scratch, took it to become a leader in electro-optics. I joined Marvell as part of the Inphi acquisition. Love Marvell, love my role at Marvell. Really excited to be here to talk to you about how Marvell is leading the electro-optics revolution. Cloud customers depend on Marvell electro-optics products to move massive amount of data inside data centers and between data centers. Electro-optics represents a huge opportunity for Marvell. Our highly differentiated product portfolio and leading market share position will drive long-term sustainable growth. So why is electro-optics? Simple. Light travels further and consumes less energy in fiber-optics than electrical data in copper cable. So all cloud data centers are wired up with fiber-optic cables, miles and miles of fiber-optic cables to connect servers, switches, storage. In fact, all links greater than a few meters today are electro-optics. The cloud data center bandwidth, it's a huge opportunity. Cloud operators deploy the latest and greatest electro-optics solutions to provide that bandwidth. In the early days, 10 gig was the best choice. That served the industry for over 10 years, 40 gig were deployed in 2013, then 100 gig in 2016. Beyond that, it was very challenging. So when you think about it, the cloud data center bandwidth represents a huge opportunity that is not met by the industry. More bandwidth drive more of everything, more compute, more networking, more storage, more data centers, more electro-optics, it's a positive feedback cycle. Marvell electro-optic SAM is growing very fast. 5 years ago, it was only about $260 million. And the cloud portion of that SAM was only less than 20%. Today, that SAM has grown to $1 billion, and it is projected to grow to $2 billion by 2024. And more importantly, the bigger part of the electro-optics SAM for Marvell is now completely flipped. 65% of Marvell electro-optic SAM is now with cloud data centers. Since electro-optics is such an important technology, so let's take a look at what is inside electro-optics and what are the solutions that Marvell provided inside the electro-optic modules. Electro-optics module converts digital data to light and vice versa. So to do that, the module needs a laser and a laser driver in a transmitter and a receiver, a detector and a transimpedance amplifier for the receiver. As the data rate goes up, a clock and data recovery was added to the signal chain to improve the signal integrity. This architecture served the industry well for over 1 decade. But at 100 gig, the analog architecture is running out of steam. It's very hard to scale analog beyond 100 gig. And that's when Marvell came in. We invented a new device called PAM4 DSP. DSP stands for digital signal processing. The secret sauce is in the core of the DSP. That's where all the magic happens. But DSP operates in the digital domain, whereas optics operates in analog domain. So for the DSP to interface with the optics, we need to develop very, very high-speed analog to digital converters and digital to analog convertors. These are very challenging technology, mixed-signal technology that require world-class mixed signal engineers, advanced CMOS nodes, analog vendors who use to build clock and data recovery just don't have that technology. In addition, the PAM DSP also needs very high-speed SerDes to interface with the host interface. And that also require very highly scaled mixed-signal analog and advanced CMOS. Taking one by one, you can find vendors who have one of these IPs. But taking all together, PAM DSP, high-speed ADC/DAC, SerDes, Marvell was the only vendor who had all of those technologies together under one roof. The benefit to the customer is, the PAM4 DSP doubles the bandwidth of the module. Furthermore, because the PAM DSP is a very powerful engine, it allows for better scaling to higher speeds. In short, PAM4 DSP accelerates the bandwidth growth in cloud data centers. Now this is the curve that I showed you earlier. 10 gig, 40 gig, 100 gig based on analog. Marvell DSP took this curve to new heights, 200, 400 and 800. We were first in every generation of PAM4 DSP. We were first in 200 gig, first in 400 gig, first in 800 gig. Each cloud has unique requirements, unique architectures. So each of our PAM4 product is designed specifically for a cloud customer. In fact, we didn't come up with a 200-gig PAM. Our first PAM chip was a 100-gig PAM, and we demonstrated at the European Optical Conference in Valencia, Spain. A lot of people came by and looked at this new PAM4 DSP. Everybody was interested. But at the end, one particular cloud customer told us, "You are too late for 100 gig, it's over. But I really like your technology. I like your team. If you got to spin this chip to 200 gig and reduce power by doing it in 16-nanometer instead of 28, we would consider select you." So that evening, two other colleagues and I at dinner, we scribbled together a plan on a piece of napkin, and we had our 200-gig PAM proposals. We came back home, presented the proposal to the customers and got selected. Meanwhile, meanwhile, the entire industry decided that they're going to skip the 200-gig generation and go straight for 400 gig. But we made the decision to stay with the cloud customer who selected us for 200 gig. Everybody told us we were wrong. But we were absolutely right. The cloud provider who selected us for 200 gig, deployed PAM ahead of everybody else. The experience that we learned from doing that 200-gig PAM chip was so valuable that we took it to the next cloud customers and worked with that customer on the 400 gig PAM chip. Although we were not the first to sample 400-gig PAM, but we were the first to be qualified, the first to go to production. And today, the majority of the 400 -- 100-gig PAM DSP that are being deployed today are made by Marvell. We show you Marvell leads the market in PAM DSP. We lead in technology, product, market position. PAM4 DSP represents a massive, massive growth opportunity ahead for Marvell. Marvell leads the market. Today, only 2 Tier 1 clouds are in full deployment, are sourcing these PAM4 DSP in full production mode. The rest of the market will follow. Next, I would like to take the opportunity to discuss what's happening outside the data centers. Cloud computing gave rise to a new type of data centers, the regional data centers. As cloud computing became ubiquitous, customers demand high availability, redundancy, quality of service. So cloud data centers need to be located in major population area, closer to the people, closer to the customers. So cloud providers realized that. And so they began to deploy data centers in metro area. But it is more difficult to deploy huge data center, 4 or 5 football field big in the middle of San Francisco or L.A. So what the cloud operators decided to do is to build data centers, smaller data centers, separate by a distance of roughly 100 kilometers, and use high-speed fiber optic links, massive bandwidth to connect these data centers together. And that give rise to huge bandwidth demand growth for between data centers at about 100-kilometer distance. That 100-kilometer distance is often referred to by the industry as a ZR distance. This new architecture drive the demand for ZR optics. Next, I will tell you how Marvell led the disruptive transition to ZR optics. In the early days, the only technology that was available to connect data centers to data centers at distances of 100 kilometers was a transport box. These are telecom-grade equipment, very powerful but big, power hungry and expensive. These boxes can drive data over thousands of kilometers. The cloud operators do not want such an equipment. They can't afford to pay for those equipment and provide massive amount of bandwidth for their regional data centers. So the industry came up with a better solution, what is called the DCI box. DCI stands for data center interconnect. It's a smaller box, a slimmed down version of the transport box. But it is still a box. So it's still expensive. So this is public knowledge. Back in 2014, Microsoft wanted to deploy regional data centers, and they did not want any of the boxes. They wanted a class -- a new class of optics called ZR optics. What it is, is pluggable modules in small form factors, same form factors what they use inside data centers, but go 100 kilometers. Microsoft went to talk to everybody in the industry. All of the OEMs, all of the module vendors and everyone said, Mr. Microsoft, it was impossible. But Microsoft did not give up. They went and talked to everybody at conferences. They talked to Marvell. At the time, I was -- we were Inphi, of course. We put together a world-class team of DSP, optics, broadband analog, and we came up with a solution. It was -- turned out, it was a cloud-optimized solution for regional data center for Microsoft data center specifically. We branded this product ColorZ. At OFC 2016, Microsoft announced at the executive forum that they have picked Inphi as a vendor for this project. The whole room of 300 or 400 executives was just stunned. You could hear complete silence. You could actually hear the pin drop. No one believed that it could be done, but we did it. ColorZ went on to become one of Inphi's most successful products for many years. And today, ColorZ is what provides the ZR connection for all of Microsoft Azure regional data centers. Forward to today, the next generation of ColorZ is 400-gig ZR. This time, 400-gig ZR is widely supported by the industries, by all top 4 cloud providers. And I'm very pleased to announce today that Marvell is the first company who took this innovative product to production. Again, here, we show you Marvell leads the electro-optics. We lead in technology, product, market position. 400-gig ZR represents a massive growth opportunity ahead for Marvell. Among all of the top cloud providers, only one cloud deployed the previous generation, ColorZ. 400-gig ZR will see broad adoption of all of the top cloud providers, plus many of the other carriers as well. Marvell leads the 400-gig ZR market. We are first to production. It's a huge opportunity to go ahead. So let's take a look at this 400-gig ZR and see what's the difference. Again, module requires laser. But other than that, every single key components inside the ZR optics are made -- are designed and made by Marvell. We own every single piece of critical component inside. The silicon photonics that make up all the optical functions other than the laser were integrated in a single silicon chip. TIA and driver are something that we always develop in-house. And Coherent DSP is a very powerful DSP engine. Coherent DSP delivers higher performance, longer reach than PAM. In fact, 4x more bandwidth per module, and 6x more bandwidth per fiber. Because of that, Coherent DSP enables wide adoption of ZR optics. Although Marvell develop and sell complete 400-gig ZR optics, we also make available as merchant DSP for our module ecosystem partners. The reason for it is cloud customers want multiple sources. They want interop. And so by enable our ecosystem partners with the same DSP, we make it easy for the end users to adapt the entire data center with Marvell DSP inside. The DSP also has extended reach for metro and long haul, which Marvell doesn't make modules. So it's great for our module partners and expand the addressable market for Marvell. The Coherent DSP Canopus, which Marvell brought to market first is now the leading market share merchant Coherent DSP. Next, I want to discuss with you a really, really important technology called silicon photonics. We built this technology organically from scratch back in the days at Inphi. The key advantage of silicon photonics is that it enables a very high level of integration, far more than the traditional discrete optics that are made by traditional optic vendors. And such an integration will be super important as we continue to scale the bandwidth to 800 gig to 1.6T to 3.2Ts because discrete optics are bulky. They cannot be packaged together in a small form factor. And at Marvell, we proved that with ColorZ. We proved that with the 400-gig ZR. You need silicon photonics to be able to put all the optic functions in a very small pluggable module. The packaging of DSP analog and silicon photonic optics together also reduce power and improve signal integrity. So this is a foundational technology for the electro-optics industry and for Marvell moving forward. And we have shown you Marvell has a leadership position in silicon photonics. And we are leading the next chapter of silicon photonics, which is integrated optics. At Marvell, we don't just build 1 single chip product, we build platforms. Our unique platform solution creates the largest ecosystem partners that consists of cloud providers, system OEMs, module, partners. This whole ecosystem needs to work together to bring products to market. Marvell chips are designed to work together really well in a platform. So customers who use Marvell platform can be assured that they can scale, they can develop their products and take their products to market sooner. The cloud end customers can be assured that with multiple module vendors using the same Marvell DSP inside, they can have multi-sources, assure enough supply, interop and also very important, scale to volume. Cloud customers when they deploy electro-optics, they want millions of pieces from the get-go. You need multiple sources and a strong ecosystem to support that. So having a strong ecosystem and a unique platform create incredibly high barrier to entry. Also, the more Marvell chips have inside the platform, the higher content [indiscernible]. So up to this point, I have discussed with you all the technologies that go inside electro-optics pluggable. As we look into the future, we see that the bandwidth -- the cloud bandwidth demand continues to accelerate. To me, the huge growth in cloud data center bandwidth demand need to work on reducing the power of the entire system. Power to the data center will ultimately be the limiting factor of how much bandwidth cloud providers can deploy inside data centers. So we see a future where electro-optics are being co-packaged together with system on chips, switches, cloud optimized silicon, compute. Co-packaged optics reduce system power consumption up to 30%, and that's a huge, huge number. So in closing, I told you, Marvell is leading the electro-optics revolution. There's a huge bandwidth demand inside and between cloud data centers. Marvell is a leader for electro-optics. We lead in technology, product, market position. We see massive opportunity ahead of us for PAM DSP and ZR optics. Thank you. Thank you very much. Thank you. And next, I would like to introduce Dan.

Thank you, Loi. Hello, everyone. I'm Dan. I joined Marvell in 2016 as part of the new leadership team, and I've been running the storage business now for 5 years. Now when I say Marvell storage, sustainable growth is probably not the first thing that pops into your head. And when I took over the storage business 5 years ago, that was certainly one of my biggest challenges. Now fast forward to today, and I'm excited to share with you one of the greatest turnaround stories you've likely never heard. And it begins with our pivot to the data center. The data center is the best market in storage. It's big, it's growing. There's high barriers to entry and our customers want to differentiate with customization. Now along the way, we built a custom preamp business from the ground up. We pioneered the DIY SSD, we assembled a world-class firmware team to help our customers customize, and we've leveraged Marvell's advanced silicon platform to move 2 generations ahead of our competitors. Now when I'm finished today, you'll be able to takeaway one very important point. And that's it, we expect Marvell's stores to deliver sustainable, high-value growth driven by the data center. Now there's no surprise that there's a lot of data being generated every year. In fact that's pushing the amount of capacity being stored in the data infrastructure or the data centers to increase at greater than 30% per year. That's a very good thing for Marvell because that means more HDDs and more SSDs. Now in addition to more drives being shipped into the data centers, there's actually trends that are driving the value of these drives and therefore, the value of our content to increase. The first trend is they need to put more capacity in every drive; second, they need these drives to be higher performance so they can access this data faster; and last, there are unique requirements for the data center, and not just the data center as a whole, but for each of the individual cloud data centers have their own requirements. So for HDD, in order to add more capacity, you primarily do this in 2 ways. You either add more platters. And every time you add a platter into an HDD, you need to add 2 preamp channels. The other way is to actually increase the amount of data stored per platter, which means you have to have smaller and smaller grain sizes, so they use a technology called energy assist. You may have heard a HAMR or MAMR used. And when you use these techniques, the HDD controller requires innovation, it gets bigger, it's more complicated. Additionally, the preamplifier now needs to add circuitry to support the energy assist elements. To improve performance in an HDD, it's really about dual actuators. That means they actually put 2 HDD controllers per drive, and these controllers need to talk to each other, which increases the complexity of these controllers. And then last, as we disaggregate storage in the cloud, our customers want to use a single unified interface, and that happens to the NVMe. So we're now starting to see HDD vendors adding NVMe into their hard drives. And of course, this requires, again, more circuitry on our HDD controllers as well as now firmware support. Now for SSD, to add more capacity, you're adding more NAND die. And more NAND die means more NAND channels on our controller and you need more throughput in a wider pipe. Now in order to get faster performance, it's all about the host interface. In this case, we're looking at PCIe Gen5 and Gen6. These are twice and 4x a throughput as a PCIe Gen4 interface. Last, to optimize for the cloud, you need to think about the fact that you have multiple customers accessing these drives at the same times. And our cloud vendors have SLAs in place that they need to make sure that this data is both secure and that they manage this access to these customers. Now, this means everything is more complex. The value of our content goes up, and we actually see generational ASP uplifts in the data center. So let's combine these 2 trends, more drives, higher ASPs. And you see a data center SAM that's going from just over $2 billion to over $3 billion, a 14% CAGR rate. And if you look closer, you'll see that this growth is really being driven by the cloud. And if you look closer, a little more closer at least, you'll see that it's no surprise that our flash-based products, the market for them is increasing in double digits. But it may be surprising to see that the market for HDDs in the cloud is also increasing at double digits. Both flash and HDD are driving this growth. It may be time to rethink HDD. In fact, as the nearline drives become a bigger and bigger portion of the overall market, it's offsetting now the PC segment where SSDs are being or replacing HDDs and we're actually getting a stable HDD market. Combine that with the preamplifier business I spoke to in the beginning, this is a business that 5 years ago was $0 for Marvell. Last year, we did just under $20 million. We are targeting $100 million next year. Now when you combine these 2 aspects, Marvell's HDD business is growing again. In fact, it grew this year. So let's pivot to SSD for a moment, and particularly the cloud -- SSD for the cloud. We are uniquely positioned for the cloud and SSD. We have 25 years of experience in storage. We have the most comprehensive IP portfolio in storage out there. We have central engineering groups that are developing PHYs faster and better than anyone else in their customized for our storage. In addition, we have access to all of Marvell's IPs, and that will be very important, and I'll share with you in a few slides why. Next, we get to leverage Marvell's advanced silicon platform. We get to be 2 generations ahead of our competition. And we get to deliver lower power and higher performance solutions to our customers. Next, we have a flexible business model, very customer-centric. We can deliver cloud optimized solutions. We can deliver custom solutions all the way to merchant solutions and anything in between. We also pioneered the DIY SSD model which actually strengthened our relationship with the NAND vendors. Because the NAND vendors want to be part of our DIY infrastructure or ecosystem, we actually get access to the latest NAND before others. And last, we truly built a world-class firmware team that enables our customers to customize their solutions. We don't deliver this one single firmware solution to all of our customers. This is not the PC market. We need to customize for our customers. So let's revisit the DIY model for a moment on SSD. Many of you know this market very well. But to remind others who may not be aware, our customers start with a Marvell controller, and we customize the firmware around their application. Marvell then qualifies and tunes for multiple NAND vendors for the end solution. Now there are many benefits to the customer in this model. First of all, it's optimized for their application and their use case. They get real-time telemetry, what's happening in their data center or in their system. They also get a flexible supply chain. They can source multiple NANDs now. You cannot do this with an off-the-shelf black box SSD. It's just not possible. And I'm really excited to say that today, I can tell you, we are now in high volume production with both our gaming partner and our cloud partner. This customization of the firmware around Marvell's architecture, our chip architecture and application optimization is extremely sticky. And I'm excited to say today, it's now also multigenerational. Let's talk technology leadership for a moment. Marvell recently announced our Bravera SC5 data center SSD family. Now this is the industry's first PCIe Gen5 product. It's optimized for the cloud. In fact, it was defined with the cloud vendors to achieve higher capacity in a smaller form factor, higher throughput, higher performance, and it has acceleration blocks and custom firmware to address all the unique needs of our data center customers. We have tremendous traction with this product already. We've secured 2 NAND vendor design wins, and we're engaged with 3 additional NAND vendor customers already. But we didn't stop there. In the last earnings call, Matt announced that we had won a PCI Gen6 custom project with a NAND vendor, and it uses our 5-nanometer platform. Well, today, I'm here to share with you that we've now put into design our first merchant solution for PCI Gen6, the Bravera SC6. And this also leverages Marvell's 5-nanometer platform. You can see why I am so confident that we believe we can grow twice as fast as a cloud SSD market. Now so far, I've talked about our HDDs and our SSDs, all of our traditional storage components. But there's an additional opportunity for storage in the cloud. Your traditional server is going to have a CPU, memory, a nick and of course, your storage, your HDD and the SSD. Now you may recall Matt mentioning cloud optimized silicon earlier. You may have heard or remember Raghib talking about the concept of storage accelerators. Well, offloading the CPU to more efficiently manage a data center infrastructure also applies to the storage infrastructure. In fact, we are now seeing the need for storage acceleration and offloading the CPU for areas such as managing a virtualized environment, securing the data, accessing additional memory or accelerating applications and even scaling out storage over Ethernet. Well, the cloud vendors are turning to Marvell for these solutions because we have the best technology. We have custom firmware. We have access to the latest advanced geometries for silicon. And we have all of the IP across Marvell, whether it be storage, security, processing or networking. And guess what? This is happening now. In fact, we've already secured our first custom design win for these cloud storage accelerators. And there's more to come. We are actively engaged on multiple other opportunities. We expect this market to grow to be over $500 million in the next 5 years. So let's bring it all back around to where we started. When we talk about our storage outlook, this is our HDD, it's our flash-based products and our Fiber channel products. We have pivoted our business. Last year, over 60% of our revenue was in the data center. We expect to outgrow the data center market with our data center solutions. And we expect to deliver sustainable growth for our entire storage business from 6% to 8%. In fact, this year, we're growing much more than that. And last, there's another growth -- long-term growth vector out there in these custom storage accelerators. Last, Marvell storage is back. And I hope that going forward, the next time someone says Marvell storage, the first thing that comes to mind will be sustainable growth. Thank you. I'll now welcome Nariman on to the stage.

Hello, ladies and gentlemen. Since this is my first time presenting in a Marvell Analyst Meeting, I thought to start by introduction. I came to Marvell from the Inphi acquisition. I manage the networking line of switching business, automotive and also the coherent DSP. Coherent DSP is used for long-reach optics. I'm very excited about the prospect of these 3 businesses. Also, I'm excited about working with the Marvell team to innovate and excel in these areas. This is my background for myself. I'm no stranger in networking industry. I was the first employee that was hired on at Broadcom back when Broadcom was a startup. They hired me to start the networking group. At that time, Broadcom was only 12 people. I had various roles at Broadcom, including managing large engineering organizations, managing businesses. And for many years, I was the General Manager of the whole networking portfolio. After 17 years, I became the CEO of a startup company in Irvine called ClariPhy Communications. The focus was coherent optics technology. After few years, we successfully sold that start-up to Inphi, and that's how I started working with Loi and Ford working on optoelectronics around our solutions and the existing Inphi products, and build Inphi into a powerhouse in optoelectronic platform. So now let me start my presentation. I'll start with the question. What's more important, processing the data or connecting the data? Well, it turns out that they are both important. Fast computer clusters cannot be built without fast networks. The faster the CPU gets, the faster the network must get to connect the CPUs. At Marvell, our mission is to be first to fastest in our networking ICs. We have a diversified portfolio. We are well positioned for 20-plus CAGR growth for the next few years. And in my presentation, I would like to discuss our strategy of how we plan to execute on these great opportunities ahead of us. The markets we serve each have its own distinct product requirements. And from the revenue perspective, we have the benefit of getting the revenue from 4 distinct market that's diversified. Cloud data centers. These customers want the highest bandwidth. They are demanding the networking IC providers to increase the capacity by 2x every 2 or 3 years at around the same power dissipation. This market is big and it's getting bigger. And Marvell is investing heavy in this market to broaden its portfolio. Now let's talk about enterprise market. This market is actually has currently a tailwind in spending. Marvell is well positioned in this market, is well positioned with the Prestera switching product families and Alaska physical layer families. In this market, we're growing revenue, and we are gaining market share. The type of switches that's required for this market needs to be feature-rich and flexible. Moving on to the carrier market. It's all about programmability because these chips go into networks that have to interoperate with the old networks but also have to be future-proofed for the upcoming standards. And now with inclusion of Inphi's coherent optics asset that's applicable to long distance optics, now Marvell has a bigger footprint in the carrier market. And the fourth one is the automotive market is very exciting. This market is estimated to be the next $1 billion Ethernet market for the IC industry. This is a market that -- the chips have to be particularly designed for auto -- automotive grade and the manufacturing and the process has to be auto-grade quality, but it's a very fast-growing market for the type of products we have. Now I'm going to talk about how our product portfolios are suited to address all these markets. Starting from the top left, is the whole electrooptics platform that Loi eloquently discussed before me. This is the combination of DSP technology for PAM coherent, the best-in-class TIA and drivers and silicon photonic technologies. It's essential technologies for optical interconnects for inside data center and for outside data center. Raghib will talk about the role of optimized cloud CPUs that are needed for specific workloads. What I'm going to talk about in this presentation is switch and PHYs, which are the fundamental building block for networks and our progress in automotive Ethernet. Collection of all these products gives us the ability to have an end-to-end networking portfolio. So what are the switches and PHYs? Where are they used? Well, they're used in switch platforms? That chip in the middle is a chip that Marvell develops and sells. It's the brain of the switch platforms. Its function is to connect the endpoints. The endpoints could be servers, could the AI clusters, could be storage arrays. The endpoints get connected through those optical transceivers that are in the bottom. What's in those optical transceivers of the optoelectronic products that I showed you in the previous slides, that gives us to innovate and offer these solutions end-to-end to our customers. Switches is what makes the network. It's what builds the network. It does the function of switching and routing and all the traffics and provide the hooks for manageability that's needed to manage a network and build the data center and enterprise networks. These devices are very, very difficult to develop. In fact, the most complex SoCs our industry does. In every generation, our engineers have to break new boundaries. And typically, the die size of the switches maximizes the allowable reticle die size in the reticle of any new CMOS process geometry. To execute, we have to have a broad IP and great engineering team, and we're very suited for that. So within the switching market, the fastest-growing segment of the switching IC market is data center. According to 650 Group, this market is about $1 billion of silicon SAM today and growing very rapidly to $1.7 billion by 2024. Today, Marvell does not have a lot of participation in this market segment. That's why I'm so pleased that we are able to announce today that we closed the acquisition of Innovium. Innovium was a very focused private company, focused on cloud optimized solution -- switch solution for data center application. Innovium chip is very high bandwidth. It has the lowest latency and also very card optimized for the application. The Innovium team did a fantastic job. Not only they prove their technology, but they want some major designs. And in fact, they won a major Tier 1 cloud operator that has deployed their technology in live networks. When we announced Innovium, we announced that Innovium brings to Marvell, incremental $150 million of business next year, and that revenue is covered by backlog. When we also did the due diligence of the Innovium team, we were impressed by the quality of the engineers, the quality of the architecture team, the quality of the implementation team. And we were also impressed by how well that architecture can scale to the future needs for our data center customers. So with inclusion of Innovium now as part of the private portfolio, that brings us the high bandwidth switch that's cloud-optimized, with the addition that Inphi asset brought into Marvell, which just happened spring of this year, and the Inphi asset is really the gold standard of optical DSP that's needed for optical connectivity inside data center and outside. With the existing widely proven and established Prestera feature-rich switch and Alaska PHY, now Marvell has a complete portfolio to innovate and address what the data center and cloud customer needs are. It's interesting, when we talk to the architects of these mega cloud data centers, they typically ask us, "Why we need a high bandwidth switch for this application?" Well, we got that now covered with Innovium. We need fast optical connectivity but a low power and efficient. Well, we got that from the Inphi now. And if they're not using optical connections, they are using copper connections, and that gets addressed by the well-established Alaska PHY. And how they connect a variety of high-bandwidth switches together and to other data centers is its feature and higher-end switches and that gets satisfied by the Prestera product family that Marvell currently has. So that gives us a very unique position to be able to innovate and provide the right solutions for the customers. And also compared to last year -- comparing last year to this year, the addition of all these product portfolios is absolutely a huge game-changing portfolio boost for Marvell, and puts us in a great position to execute. Now also the customers after we announced the acquisition of Innovium called us, and we were very pleased about what we have done. And commonly, we heard that the Innovium team, although they have done great, but the feedback that we got from all the customers is they're really going to flourish even more within the Marvell infrastructure. Why? Because Marvell has the capacity to invest more, leading edge in access to the best-in-class CMOS technology and the broadest set of IP to develop these kind of searches. So we are very excited about what the company has now had in terms of the product portfolios to address what the customer needs are. So in general, our focus is expand our portfolio in data center. We'll do it through innovation. With those innovation not just at the switching level but at the platform level. We pretty much have all the necessary IPs as required to do what the customers want. And given the bigger scale that the company has, now we have more investment capacity and gives us the opportunity to increase the velocity of product refresh cycle and keep up with the explosive demand of data that our customers have in this market segment. So now let me move on to the enterprise market. If there are 2 messages that I want you guys to take away is that the revenue is growing very nicely year-on-year and we are gaining market share. When you think about enterprise, sometimes you think about these big large buildings with wiring closets in the middle, there's the switches go on the wiring clauses and computers get connected with cubicles and offices into this wiring closet, and that's enterprise. Well, that's not the enterprise anymore. Enterprise needs to connect to outside to cloud, to hybrid cloud, to public cloud. Some of the applications in enterprise now is comingled. Enterprise has to provide the same service to those of us at home as we were in the office and enterprise is connecting the whole world now. So to make that happen, the data has to be secure, encrypted. The networks have to provide the right level of intelligence, so IT managers can manage the data, can make it fall tolerant. The switching chips and all the hardware has to be a lot more intelligent than they used to be. And that is creating an uplift in spending and giving opportunity for networking IC providers to increase the dollar content report. There is yet another catalyst for the growth in enterprise for the IC market, which is the adoption of faster Wi-Fi access points. Today, gigabit Ethernet is the standard that's used in the current generation of Wi-Fi 5, but Wi-Fi 6 is rolling out right now and gigabit Ethernet connections at the switch has to be upgraded to 2.5 to 5 gig to keep up with it. And that trend will continue in the future. This is the long-term trend. So the faster the wireless access port, the faster the switching silicon and faster the physical layer silicon. And that's why that 650 Group is predicting that by calendar year 2025, there will be 5x number of ports of connections that are faster than gigabits in this market. And again, as an investor, the key takeaway is that faster ports correspond to faster dollar content per port and higher revenue. Also, Marvell business for enterprise is very unique. This is gaining market share. It's gaining market share because it has better products. Our products are more intelligent. They're more secure. They have higher speed. We've done a great job with this product portfolio and we won all our designs. And one way to show how well we've done is by looking at the number of ports that we forecasted to ship in 2021 and compare that to the number of switching ports that we shipped in 2019. This is the internal data that come from our own manufacturing. And that number is 2x. It's 2x the number of ports in a market that at best only grew single digits in this time frame. So clearly shows that we have gained market share, and that has been showing up in our -- in the revenue growth. So in summary, the enterprise business for us is a market share -- gaining market share story, increasing content per port. We've done it, and we'll continue to do it. In fact, when I look at the trajectory of the design wins that we have won it hasn't gone to production yet, and the new products that we are refreshing, I'm very confident that we'll continue to gain market share and continue to add more content to our products and continue to improve revenue. And we have shown you that we have improved the revenue year-on-year, and we'll continue to do that. So now let's start talking about automotive, which is the next $1 billion Ethernet IC market. There are some mega trends in automotive. The first one, which I'm sure you hear a lot about is autonomous. The autonomous vehicles need to have a lot of cameras, a lot of sensors, a lot of CPUs. And all these elements need to be connected seamlessly with a network with the right quality of service. Cars also have to be connected not to inside, but to outside to 5G to Wi-Fi, but also cars need to communicate with other cars. All of that means a lot of bandwidth. And more importantly, cars from a networking perspective, need to be secure. Because imagine if you have autonomous vehicle, you don't want hacker to take over your car. So technology inside the car not only inside, but how the car connects to cloud has to be secure. And also, believe it or not, the car is not going -- is being asked to be software defined, which means you sell a platform, you sell a car. And as time goes by, you can do a software upgrade and get more features in the car. All of these trends means that there has to be a solid network inside the car that achieves the quality of service and the bandwidth that's needed. And Ethernet, it is the best fabric for that. When you think about why Ethernet? Well, because Ethernet has been around for over 30 years has evolved, has evolved into so many other networks. And this industry, rather than creating a new network might as well take advantage of Ethernet. It's available. A lot of the development to work with it. A lot of the chip providers know how to deal with it. And it's the most natural fabric for inside the car. Now looking back at the old cars, it's -- a lot of these elements are connected point-to-point, they're rigid and expensive. The new platforms, the cars that has network inside the cars, all these elements that are shown here, which are different colors are called of the sensors and actuators and other elements that need to connect to a network, displays, cameras. And all those green lines that are in the middle with all those points are the Ethernet fabric. So you can see that there could be a lot of points that need Ethernet in the cars. So this architecture makes to be secure because Ethernet has a lot of good security elements, a scalable and intelligent. And as I said, Ethernet is the future for the cars. To kind of show that trend in numbers, first of all, auto industry doesn't adopt a new technology overnight. It takes years. I remember the first Ethernet port that was shipped into autos back in 2008-2009 time frame. It took about 10 years for the industry to buy and deploy 100 million units. It's just a big number. But this calendar year, the expectation is 400 million units of Ethernet shipping into cars. It's a big number, and it's getting bigger. The forecast is by 2024, there will be about 1 billion units of Ethernet ports in the car. That's a large number. And not only is the large number that is growing, but there is a trend from 100 megabit Ethernet to gigabit multi-gig and faster 10-gig and beyond, which actually bodes well for companies like us that has very advanced technology that's always pushing the envelope and getting to that faster nodes ahead of our competition. So then you would ask yourself, why does car need so much bandwidth inside the car? Why is there so much of gigabits of Ethernet and 10 gig, that's a lot of bandwidth. But if you think about cameras, each camera can generate multi-gigabit of traffic, each camera and many cars have 5 camera or 7 cameras with sensors and connections to 5G. So there's definitely a lot of traffic that needs to get rather around inside the car. And also as an investor, the key takeaway is the higher the speed, the higher the value of the connection, the higher the dollar content and the higher the revenue. I wanted to show you some of the examples of the design win content that you see out there. There's 2 extremes that I'm showing you: One is the high content car on the right. These are cars that are brand-new architectures that are being designed primarily for electric vehicles, autonomous, these are cars that are designed with new generation of Ethernet and fabric and scalability in mind, they are future-proof cars. And they can have up to $50 to $70 worth of Ethernet content in every car. And then on the other spectrum, like high-volume older generation cars, but they still have $5 to $8.5 of Ethernet content. And we see that content to grow over time because some of these old architectures when a new car comes out, they just change 1 ECUs at a time. So it takes time for them to adopt. When you take any of these numbers and you multiply by approximately 100 million cars that get manufactured on an annual basis is still a big number. If you are wondering which car manufacturer has adopted Ethernet, it's hard to find one that hasn't. You may want to take a look at this chart and see if the car you're driving shows up on this chart. And if it doesn't, I suggest you sell your car and buy one of these cars that has Ethernet in it. Also, Marvell is the fastest-growing automotive Ethernet company for 2 reasons: One has invested heavily in auto quality manufacturing, in auto-grade qualification, and dedicated team that's developing specific Ethernet products for auto. And also that's not just -- that's by itself is great, but it's not doesn't get all the design wins. You still have to innovate. But the team at Marvell has also innovated, has been able to be first with the automotive secure switch. And as I showed you earlier on, security is very important in the car. Also, the Marvell has been able to be first with gigabit Ethernet. And the trend we see is a fast move from 100 megabit to gigabit Ethernet, and we feel very well positioned to execute and have a high market share in gigabit Ethernet market. Also the first company that delivered a sample 10-gigabit Ethernet for the cars. Based on that great execution and solid manufacturing and focus on quality, Marvell has won 28 OEMs and 7 out of 10 OEMs are now buying Marvell cars -- Marvell chips for their cars for production. And we're working on the rest of the 3. The great design win momentum is also showing up in great revenue ramp. Comparing 2019 -- calendar year 2019, which is the FY '20 that we show the revenue on this chart to what we're targeting this year, which is the calendar year '21 or FY '22, the revenue for this line of business, automotive line of business is going up 7x in a market that did grow 2.5x, but we are growing faster in this market. The revenue run rate right now for this business is about $100 million annually and is growing fast. And we expect this line of business, automotive line of business to grow faster than the company average. And also, given the competitiveness of the product, and the design win momentum that we see right now, last year, we set a target -- long-term target of gaining or getting to about 1/3 of the market. Now we are confident that based on the momentum that we see we can get to above 50% market share. Now to put it all together as a recap, Marvell is well positioned for growth. It invested heavy in data center. We are winning in data center with the Innovium asset that we just acquired and with the Inphi products because they're a great set of portfolios to win in data center. Marvell is also gaining market share in the enterprise because of better products, more secure products, more intelligent products. And in fact, we're also gaining market share in automotive Ethernet. And the revenue of automotive is going to -- has a nice trajectory to grow fast in a $1 billion new Ethernet market. Thank you. [Break]

Hello once again. Now I'm going to tell you an exciting story of how Marvell has been disrupting with the optimized compute in data infrastructure. Optimized compute is becoming a large market, and it is going to be a long-term growth driver for Marvell. We have a long and proven history of optimized compute. We started in enterprise. We expanded into 5G, and now we are further expanding into cloud. In each of these markets, we have provided optimized compute, which enable additional capabilities and the possibilities for new applications and new businesses. Let's take a look at it. First optimized compute disruption was in enterprise market. Enterprises have all kind of appliances for networking, for security, for various kind of a storage application. 20 years ago, these appliances were using many small power PC cores. They were not able to meet the requirements of next level of applications. [Audio Gap] The time when Marvell acquired Cavium, it was most of Cavium's compute revenue. OCTEON got designed in all kind of appliances and enabled new possibilities and new applications. This is still a very successful product line for Marvell. We have latest OCTEON's processor in the 5-nanometer. We shared with you last year that based on our 5-nanometer OCTEON design we are getting a great traction in enterprise. We are winning multiple designs. Today, I'm pleased to announce that one of those designs going in production next year. Today, we are shipping in 9 out of the top firewall security appliances. But we are not stopping there. We are winning a lot more design and growing this business. Wireless Infrastructure was the next optimized compute disruption. 3G RAN devices and equipment were implemented using FPGAs and fixed-function ASIC. Those solutions were good for those applications, but could not keep up with the requirements of data in the 4G LTE. This is where optimized compute was needed. Once again, we identified this problem. We leverage our field program and establish high-performance OCTEON compute platform. And we introduced RAN optimize OCTEON Fusion. OCTEON platform already had high-performance multi-core compute. It had security, it has networking interfaces. We combined it with the DSPs and hardware acceleration for RAN and implemented a unique architecture, which combines CPU, DSP and hardware acceleration. This platform became very successful in 4G LTE and became the foundation of our 5G platform. Compute is the foundation of 5G RAN platform. We started with baseband processing and transport processing. And then we expanded it into a complete digital platform. Today, we have distributed unit. We have transport processor, baseband processor, switches. And then for radio unit, we have DFEs and massive MIMO processor. As we discussed last year, with the increased demand of compute in radios, they require a full processor. Let me be very clear. This is our massive MIMO, our full processor inside the radio. This is a new content for us. When we first started, our 5G platform it was all about DU, our distributed unit. Fast forward today, we have secured multiple design which has significant high volume, and it is for massive MIMO radio solution across multiple Tier 1 OEMs. Our radio designs are driving significant content growth for us in 5G. And now with the combination in PHY, we also have optical interconnect, which completes our 5G RAN platform. We are the only company that has the entire digital platform for 5G. The overall 5G business has done very well for us. We already have established position and 4 top-tier Tier 1 OEMs in 5G. These customers are ramping. In addition to these customers, we have also secured design in 3 regional 5G OEM providers, and these will contribute to our overall Marvell's 5G business. The business is growing quickly. As of last quarter, we had already achieved $400 million run rate on an annualized basis, and there's a lot more growth in front of us. We are -- in addition to that, we are also gaining market share with the -- with our customers ramping and new design coming in production. 5G is ramping but a lot of growth is still in front of us. Just to give you an idea, the 43% of the phones that are shipped are 5G-enabled. However, only 27% network have been upgraded with 5G capability. And among those networks who have been upgraded with 5G capabilities, only 11% coverage they have. So in other words, the overall 5G is still a lot of deployment and is still going to happen. And if we take the cell phone sale as a leading indicator, this is a huge growth in front of us and it is happening. 5G is a large growth driver for us and it is growing. We are gaining -- in addition, we are gaining content share as well as overall market share. There are 2 new trends that are going in the overall 5G RAN. And often, they are confused -- there are a lot of confusion about it. That is ORAN and vRAN, so let me explain today in a very simple term. ORAN is open interface. It is all about interoperability between radio unit and digital unit among various vendors. It allows operator to pick and choose radio units and digital unit from different vendors and do a mix and match of implementation in their network. vRAN, on the other hand, is much large opportunity and disruption in the market. It is about moving the fixed function distributor unit appliances into the data center, a cloud data center architecture. It is about taking advantage of the scale of data center. This is a perfect example of cloudification. So what does it mean for the silicon vendors? Well, vRAN is not just running software. Baseband processing is a complex and latency-sensitive operation. It requires a specialized compute. The ability to process baseband in cloud computing requires vRAN-optimized custom compute solutions. The secret of success in vRAN is optimized compute silicon. And this is not only vRAN, the entire brand processing depends on optimized compute, including traditional RAN as well as vRAN. Marvell is the only semiconductor company with the field-proven high-performance baseband solution capabilities. This solution, as I discussed earlier, is called OCTEON Fusion. The flexibility and the scalability of our solution allows us to adapt to any model and architecture for 5G RAN. We are using the same baseband solution that we are deploying in the traditional RAN as a foundation to implement the optimized solution for vRAN. As a result, Marvell is industry-leading solution provider for the cloud vRAN. When a cloud provider looks for a partner to deploy vRAN, they are looking for a solution which is a higher-performance vRAN solution. They are looking for a team which has expertise of baseband processing and understanding of both RAN architecture as well as cloud scale architecture. They are looking for the track record of delivering generations of product in production in the field. Guess what? Marvell has it all and that is why Marvell is the obvious choice. We have already won PHY design. For the cloud, the business model is different. These are complex card level solutions. To be clear, in this market, we are selling entire card, which means we will have higher ASPs. The next big opportunity in compute disruption is happening in cloud and it is growing at a very fast rate. Cloud needed optimized solution to implement virtualized security and networking applications. It was being done with the CPU software or the FPGAs, but it was not enough to achieve the scale needed to enable extra capabilities and application in the cloud. In other words, it was not enough to achieve the scale of cloud. Working closely with a leading cloud provider, we implemented a solution which was optimized for cloud scale. We introduced cloud-optimized DPU. OCTEON already had capabilities of the networking and security and various kind of other application processing. We use the same technology as a foundation to build cloud-optimized DPU solution. It is a server offload solution for security, networking protocol, storage and so on. DPU in cloud has been a great business for us. We are shipping today a smartNIC as well as security DPUs in volume production. We have designed in 4 Tier 1 clouds and it is adding to Marvell growth. I already talked about the vRAN solution in cloud, but in the context of cloud, it is another DPU. This is a great business for Marvell and it is going to drive growth -- further growth -- overall Marvell growth. Now I would like to talk about the next phase in cloud-optimized silicon, the next era. Matt talked about how cloud architectures are unique and they require optimized silicon. The need -- this trend is especially important for the compute solution because compute is a big part of application processing, and that is why it needs to be aligned with the workload of the applications. This additional optimized compute is the basis of the next phase of the cloud scale. It is going to enable new application, new possibilities and new businesses. Cloud scale compute requires optimized compute solutions tailored specific to the workload. Just to be clear, these are not standard Marvell products. These are specific products co-designed by the customer and Marvell. I believe Marvell is the only company which has the knowledge and expertise and the right know-how and every capabilities needed to implement cloud-optimized silicon. This is a perfect example of what I explained earlier, where we have become an extension of engineering team in the cloud. These applications are designed for various workloads for the cloud optimized according to a specific application. This is our great opportunity that differentiates us from rest of the semiconductor companies. So let me walk you through a few examples. Cloud providers are implementing their own optimized ARM-based compute. Now at a high level, all compute look similar. They all have cores, they all have caches. They all have memory interfaces, system interfaces. But when you look closely, they are not similar. There's a lot of opportunity to optimize those compute to achieve the full performance and efficiency for a cloud-scale application. We learned it through our experience. If you remember, we had ThunderX program. We are working closely with multiple cloud providers. This is when we realized to achieve the full scale of efficiency and performance, cloud providers require to optimize their solution according to their workload needs. It means 1 size does not fit all. This is when we pivoted our business from a standard product to a custom solution working with the cloud guys. So why do we win? We win not only because we have a leadership in a process node and technology but also because we have know-how and expertise of building a very high-performance multi-core ARM-based compute. And on top of that, we also have understanding of what does it take to design a hardware, which will be able to run in a most efficient way with the existing software infrastructure. Another example of cloud-optimized custom solution is AI/ML product. Just like the compute, not all AI is same. Again, at a high level, they all look same but when you look at it closely, there's a lot of opportunities to optimize it for the target workload. And I'm not only talking about the difference between inference and training. I'm talking about actual and application. For example, you can optimize the AI solution much better, much more for the natural language processing, which will be different compared to a recommendation engine. And this is why cloud providers are implementing their own inference and training solutions. This optimized solution allows cloud providers to achieve next level of scale, and it enables them to deploy new application and new businesses. So why we win here? We win not only because we have a process node in IP, but we also have decades of experience of delivering very complex multi-chip module. These AI solutions require high bandwidth, and that is why it is important to put a lot of memory along with the silicon. It requires a lot of expertise in complex packaging and this is why Marvell is very well positioned. All these custom solutions are going to enable next era of cloud computing. It is already enabling new capabilities, new services, new business model. We believe working with the cloud providers, we will be able to deliver the innovation -- innovative products. We are engaged in a lot of design. We have won several design in these areas. We talk about how Marvell is taking advantage of optimized compute requirement in the infrastructure. We have taken advantage of disruption in enterprise, 5G and cloud. And in each one of these markets, through our innovative solution, we have enabled next level of capabilities and possibilities. Now I'm going to talk to you about next big compute opportunity. This is going to be another big growth driver for Marvell, the next multibillion-dollar market and it is automotive market. So let's take a look. The future of car technology is in intelligence. This is why autonomous cars need optimized compute silicon. Car OEM are realizing that the value that they will deliver in future, in the next years to come, will be dependent on hardware and software platform for AI and computer vision. This is why each car OEM want to control it and own it. Just like a car company could not be a car company without owning engine in the past, the optimized compute solution for automotive is the new engine for cars. Anybody who's designing ADAS assisted solution today, they need to use the existing available solution like CPUs and GPUs. But these are not efficient. These are not optimized and designed for the automotive application. And more importantly, the car OEM cannot own it and control it. Marvell has been investing in a platform for an optimized compute solution for automotive. We not only have the process node and IP, but we also have the security and networking and high-end multi-core processor. But what is needed for automotive platform is an auto-grade quality. And that is where we have been investing. We have full auto-grade quality, capable product line function building block that can be used to implement optimized auto compute solution. We have been working with multiple car OEMs. And today, I'm pleased to announce that we have secured the first high-volume optimized auto compute design. This is a great opportunity for Marvell and we have a strong opportunity pipeline. And it is going to be a multiyear growth driver for Marvell. We are engaged with OEMs in every geography and the discussions are at various stages. But as you know, it takes time to ramp design and production in the automotive industry. So while it is going to be a great growth driver, it is going to take several years to ramp. Automotive compute is a new emerging market. These are early estimates of the size of the market. However, just to give you an idea, there are 80 million cars are shipped every year. Now over time, these cars will become autonomous, ADAS-enabled. And the solution, the compute you need to enable that autonomous car is what we are talking here. And I'm not talking about $10 or $20 silicon. These are multi-hundred dollar solution. And a typical car needs 2 of those because of the fact that you need for higher availability and redundancy. So as the adoption of autonomous car increase, this market is geared to high growth. This is going to be a multibillion-dollar opportunity and incremental opportunity for Marvell. So we talk about how Marvell is very well established to be the leader in optimized compute. And it has been disrupting multiple markets. We disrupted enterprise market, added value and we are an established player and it is growing. We brought a lot of value through optimized compute in 5G market. We are the only silicon provider in 5G with a complete digital platform. That business is great, it is growing and we are gaining more share with vRAN. We are very well aligned with the disruption of optimized compute in cloud. We are working very closely with our customers, innovating together, developing together and bringing new solution, which is opening up new possibilities, new application, new businesses. We are capitalizing on this huge disruption of cloud-optimized silicon. And finally, we are investing in optimized compute platform for automotive. This is going to be a massive market, and we are very well positioned with our investment as well as engagement with the customer. And that is why this is going to be a huge growth driver for Marvell. With that, I would like to invite our CFO, Jean Hu. Thank you.

Thank you, Raghib. Thank you all for joining us virtually today. As you have heard from our team this morning, the investment Marvell has consistently made to build the best technology and product portfolio to address the most exciting market opportunities are driving incredible momentum across our entire business. Today, Marvell is a fast-growing company, leading the future of data infrastructure semiconductor market. Our technology and product leadership, customer-centric business model, our talented team and the strong execution capability have driven accelerated revenue growth, earnings expansion and strong cash flow generation. More importantly, it will allow Marvell to continue to invest ahead of industry curve to drive long-term value creation for all stakeholders. At our prior Investor Day, we set up our financial road map for long-term value creation to focus on the key elements of our business model: accelerating top line revenue growth, drive earnings expansion faster than revenue growth and disciplined capital allocation. Specifically, we set our long-term revenue target in the range of 10% to 15%. We updated it to 12% to 16% post the Inphi acquisition. We also said we're going to expand earnings much faster than top line revenue growth. We showed you our scale curve to expand operating margin from 20% at that time to 35% the target. We also talked about our disciplined capital allocation approach to focus on return on investment. Looking back, our team executed very well. We did exactly what we said we're going to do. First, on revenue. As you can see from this chart, we grew our top line revenue for the last 6 consecutive quarters. Marvell organic revenue growth, represented by the blue line there, accelerated from 5% to 17% in the first half of fiscal '22. At the midpoint of our Q3 guidance, our revenue is expected to grow more than 20%, exceeding our long-term target. We also closed the Inphi acquisition in Q1 fiscal '22 within 6 months of the deal announcement. Inphi acquisition was accretive to our Q2 top line revenue growth, the fourth quarter as part of Marvell. The combined company at the middle point of our Q3 guidance is expected to grow revenue by more than 50% and reach $4.6 billion annualized revenue run rate, a historical record for Marvell. Our revenue acceleration is made possible because we have multiple drivers to drive our top line revenue growth and the $20 billion market opportunities which will continue to grow. To provide more information and transparency to investors about our key business growth drivers, we have started to report revenue by end market to align our financial reporting with our business driver and the market-focused resource allocation approach. To start with, in data center, we grew our revenue by 62%, driven by a broad set of products from storage, DPU, PAM-based optical interconnect inside the data center and adoption of optical interconnect between data centers. We are at a very early stage of many of these product revenue ramps. Our carrier infrastructure business grew 38%. In addition to Inphi's contribution to our wireline business, our growth is driven by continued 5G adoption and the new program ramps with our key customers, Samsung and Nokia. Our enterprise networking business grew 41%, driven by our enterprise networking portfolio to continue to gain market share and also the market gig adoption. Auto industrial business grew 125% year-over-year, driven by automotive Ethernet business ramp Nariman discussed earlier. As a team, we are very pleased with the revenue acceleration performance, and we strongly believe the merit of scale and operating leverage to drive earnings expansion faster than revenue. We improved our gross margin from 63% to 64.8% despite a very challenging supply chain environment. Our 2 highest growth end market, auto industrial and the data center have the highest gross margin of the company, too. We expand our operating margin by 900 basis points to above 30%. As a result, for our operating leverage, we drove our earnings per share growth by 62% in Q2 fiscal '22. And we're on track to drive continued revenue growth and earnings expansion in the second half of fiscal '22. We are very pleased with our performance. It's an exciting time to be part of Marvell because as you heard from our team this morning, we have been laser-focused on what Marvell can become in the next 5 years and how we get there. So let me bring together what you heard this morning from our team in the context of financial model to talk about our financial priorities going forward, to drive sustainable revenue growth at scale, to scale our business model to expand earnings faster than revenue growth and continue to generate strong returns for shareholders. First, on revenue growth. As you recall, Matt discussed earlier about extraordinary opportunities ahead of Marvell and how we have transformed the company through both organic investment and acquisitions to address these opportunities. Today, infrastructure represented 84% of our revenue. As many of you know, our infrastructure business tend to have a very long product cycle. The revenue transformation you see today is really a function of the investment that we have made and we talked to you about at our prior Investor Days. What you heard today, the exciting technology and the product portfolio, the cloud, the customized silicon opportunity, our complete platform to address those opportunities and the tremendous design win momentum, those are going to drive next phase of Marvell's transformation and the top line revenue growth. So let me now summarize our key revenue growth drivers going forward. It's important to note our revenue growth drivers are long term in nature and we assume a normalized economic environment. As you know, Matt talked earlier about our market opportunity to grow from $20 billion to $30 billion at a 13% CAGR. Within the market we serve, there are several secular growth trends in cloud, 5G and automotive, which are driving those markets to grow even faster at a 20% CAGR. We expect our revenue in cloud, 5G and automotive to grow even faster than the market growth rate at 2x of the market. The largest incremental revenue growth will come from cloud data center, driven by a broad set of product our team discussed today from storage, networking, electro-optics, compute and security. Based on strong revenue momentum we have and the design wins Matt discussed earlier, we are on track to drive exciting growth in this fast-growing market. Next, 5G. Raghib talked about significant design wins we have with our key co-OEMs and regional OEMs. We also extend our leadership into ORAN, vRAN market. We expect our revenue continue to ramp significantly, driven by 5G adoption globally and our new program ramps, including 5-nanometer programs. We also expect to get a benefit from the initial deployment of ORAN and vRAN. In automotive, as Nariman discussed earlier, we currently already have the run rate of $100 million. Based on our leadership products and the design wins we have, we now expect to get to 1/3 of market share much earlier than we told you last year. We also expect our new target is 50% market share. Now looking at the rest of infrastructure market. We expect our revenue to grow faster than market. The primary driver is enterprise networking business. We expect our enterprise networking business to continue to grow double digit, driven by market share gain and multi-gig upgrade cycles. We also expect our wireline business to continue to grow faster than market, driven by our coherent DSP leadership and the technology transition in metro and the long-haul market. Our on-prem business and the industrial business are expected to grow at the same rate of the market. Overall, as a result of market acceleration, acquisition of Inphi and Innovium as well as higher demand for our products, we are updating our long-term target to 15% to 20%, one of the highest growth rate in semiconductor industry. Now let me shift gears to talk about our gross margin and operating margin, how we can continue to drive earnings expansion faster than revenue growth. We are very pleased with our strong gross margin performance in Q2 fiscal '22 at 64.8%. Our 2 highest growth market, auto industrial and the data center, actually have higher than corporate average gross margin. We expect not only that we'll continue to drive top line growth from those 2 end markets, we also believe the gross margin of those 2 end markets are going to be -- continue to be accretive to our overall gross margin. Our enterprise networking gross margin is around coverage. As many of you recall, we have improved our enterprise networking gross margin from below 60% to about around corporate average. We expect to continue to improve gross margin going forward to be above corporate average. Our consumer business is going to become a smaller portion of our business, and we expect the impact going to be smaller in the long term. We'll continue to improve our carrier infrastructure gross margin. We do believe we'll benefit from the scale of revenue ramp when we started to significantly ramp the 5G revenue going forward. Overall, we believe that gross margin is a strong reflection of our IP, technology, product leadership and the value we add to our customers. We are targeting our non-GAAP gross margin to be in the range of 64% to 66%. The variation is largely due to both carrier market and the cloud data center tend to have lumpy product cycles. Now let's talk about the operating expense. We intend to continue to drive operating leverage while investing to support our long-term growth. As you can see, we have been very successful to drive scale and operating leverage. As a result of our revenue growth from below $3 billion in fiscal '21 to $4.3 billion run rate based on Q2 fiscal '22 performance, we have reduced our OpEx as a percentage of revenue from 43% to 34%. We expect to continue to grow OpEx slower than our top line revenue growth to further drive operating leverage. At the same time, given the tremendous opportunity we have as a company, we want to make sure we are investing at the right intensity to support our long-term growth. We are spending over $1.2 billion in our R&D, and our team is excellent and disciplined to drive return on invested capital. As you can see, 95% of our R&D are focused on cloud, carrier, enterprise, automotive, 3-nanometer, 5-nanometer advanced technology development. In addition, as Matt and Raghib talked about it earlier, we work with our customers, and we have long-term partnership to help our customers to develop differentiated solutions. As part of our partnership, we also got co-investment from customers, which increased our R&D scale significantly, and also, we can leverage IP across all our business and all our end markets. Overall, we are targeting non-GAAP operating expense as a percentage of revenue to be in the range of 26% to 28% of sales. Now let me summarize our financial model. As you recall last year, exactly a year ago, we set up our financial model, long-term financial model. The top line revenue growth was 10% to 15% and the operating margin was 35%. Today, we set up our operating model as top line revenue growth of 15% to 20%, gross margin in the range of 64% to 66% and operating margin in the range of 38% to 40%. This is a very compelling financial model in our industry. And as you can see from our recent financial performance and the momentum, we are growing our top line revenue above the target, and our gross margin is right in the middle of our target range, and we are improving our operating margin each quarter rapidly. So we are on track and making great progress toward our model. Now let me switch to capital allocation. Our capital allocation principles continue to be consistent and the same. First and foremost is organic investment to support the growth opportunities ahead of us. Our focus is to invest in talent and the people, as Matt discussed earlier, we are also driving the technology in 3-, 5-nanometer and the foundational IP development. We invest in supporting infrastructure, too, to continue to scale our supporting structure. Secondly, we'll continue to explore acquisition opportunities with a mindset of long-term strategic focus and strong financial discipline. During the last year, we completed 2 acquisitions, Inphi and Innovium. Both transactions are very compelling strategically and financially. As Matt discussed earlier, we now have all the pieces to execute on our road map and address opportunities ahead of us so we can generate significant returns on those investments. Finally, our business is generating strong cash flow today. As we move forward, powered our target model, we're going to generate more cash flow going forward. We are very committed to returning cash to shareholders and a target to return more than 50% of free cash flow to shareholders in the long term. We are also targeting to offset employee equity plan dilution through share buyback at a minimum. We currently intend to maintain the dividend level we have and prioritize share repurchase. In the near term, our first priority is to pay down the debt we used for Inphi acquisition by $100 million to $150 million each quarter. Our target gross leverage ratio is 2x. We do expect to deleverage quickly as our EBITDA is going to expand significantly each quarter because of the top line revenue growth and operating leverage. Once we are on track to achieve our targeted gross leverage ratio, we'll have the flexibility to start share repurchase. As a company, we do have a good record of share repurchase. We returned $1.2 billion cash to shareholders since Q3 fiscal 2017 at an average price per share of $18 and generated 36% compounded annual return. So in closing, today, Marvell is a fast-growing company, leading the data center future. And we are building one of the best financial model to drive sustainable revenue growth and deliver significant value to our shareholders. With that, I want to thank everyone for your interest in Marvell. We will now go to Q&A session.

Wow, what an exciting morning. I mean, I'm in the middle of all this and I still get excited every time I see this. And I just want to thank my whole team who worked so hard on these incredible presentations. We're missing Nariman. He actually had to run off to go secure a design win so we figured we'd make that trade-off so we can answer questions for him. But with that, I think, Ashish, let's go ahead and get started with the first question.

The first question is Matt, you guys did a great job of describing your long-term strategic growth drivers, but investors are also very curious about your take on the near-term economic environment, especially, how do you think about supply, your ability to actually continue driving growth at or even potentially above your long-term target model?

Yes, great. Well, first, I'd start off by saying business is incredibly strong. If you look at where we've been growing and even in Q3 where we guided at the midpoint, we're actually growing above the long-term model range that we set. This is not due to a COVID bump. I mean, this is due to really these fundamental growth drivers we talked about in cloud and 5G and auto and also this very strong demand we're now seeing in enterprise. I'd also say that in the last 6 to 9 months, I give our team a lot of credit, we've really done tremendous work, right, to work with our supply chain partners on telling them our story, actually working with them to plan our business together, signing long-term agreements and securing additional capacity that we see coming online next year to really help us drive our growth. And in fact, some of that capacity will -- incremental capacity will also be coming online in Q4, which is really exciting because of all the growth opportunities we have ahead of us.

The next question is it on PAM4. So Loi, can you talk about your PAM4 market share? And how do you think about the long-term competitive position for Marvell?

Great question. As you know, we invented PAM4. We were first for 200-gig -- first for 200-gig, first for 400-gig, and we're leading the market for the next-gen 800-gig. So obviously, we have a high share of the market. But it is not just about the PAM4 DSP. It's a complete platform, the TIA, the driver, the firmware, how all the pieces need to work together and the strong ecosystem partners who could ramp these things to millions of pieces from day 1. So the barrier to entry is actually incredibly high, so we expect that we will continue to maintain very high share in the PAM4 platform.

Thanks, Loi. The next question is on cloud. You explained clearly that cloud needs customized silicon. Why do you think that they want to work with Marvell versus developing these solutions internally?

Yes, I'll have Raghib answer that question.

Okay. So as I explained, it is not about a point product. It's a combination of process node leadership, IP leadership as well as product expertise and know-how leadership. As I explained during my presentation, Marvell has actually established leadership in each 1 of the components, which are capabilities which are needed to develop a leading optimized solution for cloud. So for example, we have the compute, the security, the storage, the switches and optical interconnect. So when a customer is looking to develop and optimize product which is going to combine the capabilities from multiple of these markets. They are looking for somebody who has developed those products, take it into production, harden it, ship it in various use cases, so that you have the right expertise and the right capabilities. And this is why it is about having a partnership where they develop certain portion and we develop a certain portion. It is about bringing all the capabilities where we put our architects with their architects and we are co-designing this thing, right? So this is where my -- as I said, is one of my jobs is to have that interface with the customer and really understand what is needed and then bring -- pull in architects from various BUs. So for example, recently, just yesterday, I had a meeting with a cloud provider, and I had to pull the architect from storage as well as switch as well as optical actually because the solution requires all those combination in that, right? So those kind of things happening again and again, and this is where, as I said earlier, Marvell has established itself to have this advantage compared to any other supplier out there because of the fact that we have market-leading products in each 1 of these areas.

Ashish, I would just add to that, if I could, in what Raghib said. I think part of that is also how the company culture operates. I mean, we really operate as 1 team and it's how we set our goals, we drive our incentives and the type of people we hire. This is not a siloed company where different BUs are completely stove-piped from 1 another. We actually -- the teamwork extends. Raghib's got this great new role to coordinate, but it's really a team-oriented Marvell first, 1 Marvell type of company. And I think you need that as well to all the pieces to really make this happen. And I think the customers see that.

A follow-on question on the economics of the cloud-optimized model. Historically, co-investment with customers in joint development and NRE translated to lower gross margin. But your model is showing higher gross margin, suggesting that, that paradigm is broken. So the question is, what has changed? Why is the gross margin model different going forward in cloud-optimized?

Sure. Well, I mean I'll give 1 example. It's a win-win when you can do an optimized solution for a customer, meaning they actually can save costs, improve their performance, get a better outcome. And for us to be able to do that for them, that's got value ascribed to it. And as Jean showed, I think in the bubble chart, we have a very healthy margin profile in this segment. And I think it's a testament to the value that we can deliver. And remember, we have a range of products as well that we sell into the data center, right? We sell switches and optical components and storage products. And there is a range of margin profiles within that but the overall blend is quite strong. So I wanted to dispel the notion that cloud-optimized means a 10- or 20-year-old version of what people think ASIC might be. It's the furthest thing from that actually, although there is an aspect of our business where we do full custom. But even in those engagements, there's tremendous value, as Raghib said, to really bring in the Marvell IP portfolio and figure out how to add value. So I think it's got a compelling financial model behind it, but also it's because of the value that's being delivered into these applications.

Just staying on the cloud theme, you outlined a new target for large customers, $100 million-plus customers. You took that from 17 to 19. Can you elaborate how many cloud customers are part of the new model?

Sure. Maybe just to take a quick step back. If you go back to just 1 year ago today when we had our Investor Day, pre Inphi, we actually had 13, path to 13, $100 million, customers. So what was exciting about Inphi is they brought us an additional 4 and now we said we're going to 19. And so the -- what's happened now is we've actually got an additional cloud customer that we're going to be adding as a $100 million and an additional networking customer. And so that's really -- adds on top of where we were before.

Right. So we're going to move away from cloud for a few moments. Question is on automotive. You've made rapid progress on Ethernet and you talked about entering into the compute market. Can you talk a little bit about how Marvell got here and how do you think about your prospects going forward.

Sure. I'll speak to that. when I joined the company 5 years ago, we had nothing. I mean, we didn't have any business. There was a small engineering team that had a vision that said that this could be an important technology. The group was not really getting attention, any attention inside of Marvell. And my background, prior to being CEO of Marvell, is I had extensive experience for the prior 15 years or so in automotive semiconductors, and I knew this Ethernet trend. So between myself and a few others, we really made a concerted effort to build a business from the ground up. So we installed the right quality processes and controls and people. We staffed the group. We hired a Vice President. We created its own business unit. And we've developed winning products with exceptional quality. I mean, we just had a review the other day. A few of us were there with a leading car OEM, who is one of the largest producers of cars in the world, and they were astonished on the fact that we're shipping now below 1 DPPM kind of quality levels on our Ethernet products. But we always had a vision, Ashish, that this was going to be much larger than just networking. So we've got Ethernet products today and switches and PHYs and now going to multi-gig, and Nariman showed that's super exciting. We believe we can get 50% of that market. But then beyond that, I think long term, I've always believed that the biggest next market opportunity in automotive was going to be for the data-centric IPs and that includes the compute, the storage, the networking, the security. And there's even experimental things we're looking at optics, right, in the future. So I think it's going to be a very much a greenfield for Marvell, and Raghib really talked about what is the first of, I think, many really unique opportunities across those other categories besides just networking.

Right. Next is a 2-part question. First on M&A and the second 1 is on return to shareholders. So you've done a lot of M&A over the last few years, very transformative in nature. It feels like you have everything now. So can you kind of comment on how you think about future M&A going forward? And related to that, can you discuss your priorities on returning cash to shareholders? Can you just comment on both of those, please?

Sure. I'll take the first part and I'll let Jean talk about the second. So look, I think the M&A that Marvell has done over the last 5 years has been just transformational for our company. And obviously, if we hadn't done this, I wouldn't be sitting on stage here with 2 founders, okay, of 2 of the most successful, highest-quality semiconductor companies we've ever seen, right, which is Cavium and Inphi and their teams. So hadn't done that, we wouldn't have these kind of people. Beyond that though, look, we did what we had to do, okay, to get Marvell out of the situation it was in 5 years ago. And I'm very proud of the accomplishments of the team to get the companies integrated and also the organic investments that have paid off and Dan, I mean, is a shining example of that, taking a storage business that was effectively left for dead and now creating one of the most exciting businesses inside of Marvell. So all that's gone really well. We have all the pieces now, okay? Now is the time for us to focus, execute, take advantage of all these opportunities, not only that we've already won but they keep coming in the door every week as we continue to sort of talk about our platform and engage with customers. So that's really where our focus is. Now we're very good at M&A. These are always -- we're always going to have a lookout and there are some unique promising young companies, there's technology out there that we would look at. But if you just look at the portfolio, we have all the pieces, it's the best 1 in the industry. We've got what we need and I'm really focused on driving that in the near term. And then all of that and I'll turn it to Jean as it relates to how we think about shareholder and capital returns because those things -- the 2 things go part and parcel.

Yes. I'll just follow what Matt said. If you think about Marvell, we are building one of the best compelling financial models. So today, our revenue growth is above the target of 20%, and we're generating tremendous cash flow. Going forward, when we move toward our target model, we're going to generate more free cash flow. Of course, our #1 priority is investing, but following what Matt said is we actually really -- once we get our target leverage ratio, we're going to return cash to shareholders. And the priority is share repurchase. We'll keep our current dividend level and return cash through share repurchase we can do.

Following on the M&A theme, very exciting to see you close Innovium. Can you discuss a little bit about how you feel about Innovium as part of Marvell and how you think that will compete against a primary competitor in that particular space?

Yes, sure. I'll comment on that and maybe Raghib can add as well. We worked closely together to go get that transaction completed and now Nariman is managing it. Look, what Innovium does is extremely hard. We have our own switch business, right, that grew up in Marvell. It was here 5 years ago. And I had concluded for many years that the -- it was going to be very difficult organically to try to actually go compete because we have some -- there's a very, very strong competitor in that area. What Innovium did was find a way. And it's an incredibly strong team. They managed from basically starting from nothing to being a venture-backed startup to generating significant position in the market, including a major presence with one of the Tier 1 cloud companies. And as Nariman said, we've got strong backlog for next year. We've got the demand. In fact, the demand is even higher than what we think we can ship currently so we're actually working on that. So it's a great setup in terms of the asset. But like a lot of small companies, I think there was a number of customers that played around with their parts that was always interesting but never -- was always worried, well, what's going to happen? Is it going to get acquired? Is it going to make it? And to some extent, even with Inphi and Cavium, which were established companies that were publicly traded, we saw that same effect, right? Once they became part of the Marvell platform, it opened up massive new sets of customers. So I think the same thing is going to happen. The second we announced it, all the customers that have been sampling the product move basically to how do we engage. And so I think the prospects look great. We've only owned it for a day obviously, but the integration planning has gone well. And I'm very optimistic about that business and what it can do over the next several years as part of the Marvell family. And so I think the prospects look great. We've only owned it for a day, obviously, but the integration planning has gone well. And I'm very optimistic about that business and what it can do over the next several years as part of the Marvell family.

Just following on the Ethernet theme, you showed a very impressive chart outlining 2x port gains over the last couple of years. Can you comment on your competitive position in the Ethernet market? What drove those share gains? And how do you think about it going forward?

Sure. I'll have Raghib comment on that. He was running that business and did a great job. And of course, Nariman has now taken the mantle. But maybe you want to comment, this is our enterprise switch business, Ashish. Yes?

Yes. So overall, if you look at the overall Ethernet enterprise. We have invested over the last several years, and this is where we differentiate in our product when it comes to the feature, the visibility, capability and that security and all that. And that kind of is what is needed and as the overall market is evolving into more of a borderless enterprise, the need of telemetry, the need of security increasing. And that is where we are gaining more and more share in the market. And that trend is still going on. So we have established a product, which is a market-leading product in that market. And with this trend of more feature requirement, more security requirement. We are seeing a constant trend in the growth of our market share in overall adoption of our product.

The next question is around our update on our long-term revenue growth target. Can you bridge the prior target of 12% to 16% midpoint of 14% to the new target of 15? To 20? Is Innovium a big reason for the update? Or is it broader than that?

Yes. I would say Innovium is part of that. I think it's part of the bigger shift that we've seen even in the last 12 months, to the market growth rate and opportunity really around this cloud optimized silicon framework that we talked about. So if I look at what's driving our model update, it's really that cloud portion, which is continue to build momentum, right? If you go back into, say, early 2020, we said, "Hey, for the first time, cloud was more than 10% of revenue for Marvell." And it was like, "Wow, that's a new data point and then we sort of get to the Investor Day last year. " And at that point, our cloud business was whatever it was, $400 million kind of run rate. Fast forward to now, Inphi is under the 10%, it's $1 billion. I mean, this thing is moving at an incredible rate, right? So Innovium certainly helps boost that, but it's also the prospects of the core Marvell business, the organic product lines we're investing in, plus the Inphi acquisition and I think the adoption there as well.

Just bridging on that next question is it seems like part of the increase in the growth rate was also some of the new incremental design wins you talked about. You mentioned that some of them are pulled in as well as you showed at 2x. So can you comment on what drove both the pull-in as well as the significant jump in the outer years?

Sure. Yes, I think a couple of things have happened. One is when we talked about those wins initially, they were confirmed design wins and we certainly had an outlook of -- and a view on when they would ramp up and the economic value of those. Subsequently, a couple of things have happened. One is we've concluded the contracts on those. So we've completed development agreements. We've got schedules, we've got pricing, volumes. And given the trend in the market, you got to assume that all that sort of floated up since we started. We've also won some additional design wins, right, since that time. And candidly, the customers are pulling us and my team knows this to do them even faster. So it's not like the schedules we have are exactly what everybody wants. So we're actually working at improving those. So I think it's really a testament, Ashish. The way to think about it is it's just an example of how fast this transition is moving within the market where that old framework of there was sort of a general purpose way to do it, and then there's this opportunity set around cloud optimized. I'd say my weekly one-on-ones with my BU direct reports, every week, there's new opportunities coming in. They're big, they're meaningful. They're addressing new applications. And I think our model is kind of feeding that now. I'm not sure that if you go back a year ago, they would have said, we could even feasibly do what we're thinking without a Marvell in the room. They may just have gone and used some general purpose product to go do that. And it wouldn't so I think it's only going to continue to get stronger, Ashish, in terms of the opportunities we see. And I think this is just 1 example of the momentum in our business in terms of how we view even opportunities we won, call it, 9 months ago versus where we see them today.

We're going to switch gears and now talk a couple of questions, take 2 questions on technology. So the first one is for Loi. Loi, you showed some very interesting charts on both silicon photonics as well as copackage optics. Can you elaborate on these new packaging technologies Marvell is working on, especially relative to CPO? How do you see that intersect in the road map? And then what's the benefit you see for customers?

Good question, Ashish. So we started working on silicon photonics nearly 10 years ago, actually, when we -- and then we develop the entire way to integrate, have a couple like into the PC of silicon photonic. That's a really, really secret sauce, a magic know-how that most people don't realize. How to do that in a very efficient way, putting everything together, both electronics and optics in a very small form factor like COLORZ. And then we went on to continue to do that with the 400-gig ZR, extremely complex packaging technology. And then we've continue on with that path, 2.5D, 3D packaging, chip-on-chip, heterogeneous integration, you will hear that term a lot. What it means is you -- now the technology allows Marvell to integrate optics and electronics in this way, stacking on top of one another, very compact, very high integration. And it's really a critical piece for a very high bandwidth products. And then you do that again with the DSP and the rest of the SoC. So the example that Raghib show at the beginning and then I show during my presentation, that is legally an optical -- electron optics system on a chip. It had Silicon photonic, it got lasers, it had a TIA, it had driver. It got DSP on co-packaging together in a very compact form factor. That is the basis for developing the next generation of even more compact 2.5D, 3D packaging co-packaged optics. In that case, the optical system on chip will be co-packed with the switch. And now Innovium is part of the Marvell family. And so we could do that internally. But even so, we are supporting open ecosystem. Our cloud customer would like to have an open standard for co-packaged optics. And so we are working in a way so that the co-package optic can be done in an industry agreed upon standard way as well as our internal solution. Thank you.

A broader question on process technology. So it's very exciting to see Marvell talk about introducing a 3-nanometer platform. Raghib mentioned 224 gig SerDes. So can you elaborate a little bit on timing? What's driving the quick move towards 3, Marvell just announced 5 just last year. How do you see that helping your road map going forward?

Sure. I'll just make a couple of comments, and I'll let you chime in. I think, look, the 5-nanometer platform that we announced last year where it was really visible that we had made this pivot from follower to leader. It's been a home run. So we absolutely are going to be there as a leader as well on 3, and it's critical, it's strategic, and we're investing. And Raghib, maybe you want to add some of the details around how we think about it because it's more than just the node, it's more than just a Moore's Law jump, it's actually much broader.

That's true, Matt. So first, establishing yourself the leader in the process node is not about a onetime, okay? So you have to keep leadership and we have established ourselves in the 5-nanometer. And I'm really confident we will repeat the same thing in 3 nanometer. Now talking about what is 3-nanometer, and we did a press release with TSMC as well. It's not just a node. It is about the whole platform. So Loi just mentioned about heterogeneous integration. What is it? So as you go in a latest process node, there are pieces of the silicon, you may not want I do in 3-nanometer, keeping 5-nanometer, for example. This is where you need to invest in this whole technology platform where you can bring chiplets from -- or multiple chips from different process node and put together and have all those package design and so on. So when we say about 3-nanometer investment, we talk about process node investment, SerDes IP investment. We establish ourselves leadership in 112 gig. We are investing in 224 gig to keep that leadership. So I'm pretty confident with our engineering team that we will achieve that one, too. And then also, having a leadership in putting this thing together. Having the leadership of having the platform and the package. So all those thing require multiyear kind of R&D and investment and research. And this is where we are leading to keep our leadership established.

We're going to switch gears to Jean. So the question was on OpEx. Marvell has done a tremendous job of managing OpEx over the last few years, but you're driving incredible revenue growth significant number of new design wins and a lot more excitement ahead. So how do you think about OpEx going forward? Do you think you're spending enough to sustain this level of growth you intend to drive?

Yes. So given the opportunities we have as a company, it's definitely most important to continue to invest and at the right intensity on R&D side. So our team has been super disciplined, super focused on the investment. If you look at our investment, we are very focused. As a company, we're spending about $1.2 billion on R&D annualized run rate, right? In addition, we do get the co-investment from a customer side. So we leverage all the investments to do not only product investing, but also advanced process technology investing. I think it's important that we'll continue to drive the R&D expenses going up -- going forward. But compared to our top line revenue growth rate, the investment, the OpEx growth is still going to be lower than top line revenue growth. So overall, when we ramp our revenue significantly up, you can see the continued significant leverage of our operating model.

A question for Loi. Loi, you showed a very interesting graph showing 400ZR port counts, units increasing 6x over the next few years. Marvell or Inphi really was a pioneer in introducing this technology. So how do you think about what this growth means for Marvell going forward, your market share in this particular market?

Well, as we discussed during my presentation, we invented this whole class of products from ZR Optics. We started with color for 100 gig and now we move into 400 gig. So we have a tremendous amount of know-how, how to get this product works in a cloud customer environment inside of switch. We expect that we will be the market leader for the transition to 400 gig ZR. We demonstrate that we are the first company who are taking this product to production. So we believe that we will continue to maintain market leadership in 400 gig ZR optics as it moves forward.

As fitting for today's presentation, we'll take a last question on cloud optimized silicon. So it's a long question. Let me read it. "Matt, while you've been more focused and earlier than most talking about cloud optimized silicon. Can you comment about the moats you have? To what extent the IP blocks that you have, which is not available at other competitors. What feedback do you get from your cloud customers when you engage with them on this optimized model going forward?"

Sure, Ashish. That's a great question. And I think the way I would propose that we conclude the Q&A is I'm going to ask the members of my team who are running these groups because remember, this isn't just a one size fits all. It's not like we have this 1 IP and then that's -- it's actually how you bring it all together in the solution, right? So maybe I'll start with Dan. I mean you talked about the whole cloud accelerator opportunity, the trends in storage. So why don't we go Dan and then Loi and then Raghib, and have each of you comment about in your area, how this comes together.

Yes. I mean I think we have 25 years of experience in storage, and we're organized in such a way that we have teams of architects, product refiners and really innovators within storage that are continuously driving new innovations, customized for the data center. But what's unique is that we also get to leverage the IP in the other organizations, whether it be advanced processing, networking capabilities, security. So when we go in and talk to our customers and work with our customers, we have a very unique view. We also have unique capabilities. And then you combine that with our firmware capabilities, then we allow massive customization with a very data center-focused firmware team that's optimized to customize. It really is unique. And there's no other competitor in the world that has all the capabilities that Marvell has in storage, we believe.

So in terms of the electrical optics, as you know, we invented the PAM and we are leading the market for every single generation of PAM. We invented ZR Optics, and we're leading the 400ZR Optics. So our cloud customers, they are really depend on Marvell to work with them to define what is the next generation of whether PAM or ZR Optics that they need for their cloud data center. These are very, very cloud optimized solutions. It's not a one-size-fits-all. Everything is tailored for particular customers. So there are really no other vendor who has that kind of technology, capability, products and the deep level of engagement with each of the cloud provider today as Marvell.

So as I already explained, we pioneered data-centric compute, data processing unit concept. And we have 25-plus years of experience of leading data-centric processing as well as security, right? So as I said earlier, it is not only about 1 category and 1 capability. What makes Marvel unique is Dan talk about storage, Loi talk about Optical. I just explained the compute and security. Nariman earlier presented about the switching capability and so on. So having the ability to be able to have expertise to develop those products and really gain market knowledge and technology knowledge through deploying that product in production in the market, and that is what is unique. And what our cloud customers like is this, we are not kind of keeping those knowledge or IPs close to our test only. And we are letting them to work with us, to participate in codesign, leveraging all those expertise and IPs and capabilities to develop optimized solution, leveraging capabilities from multiple areas. And that's what makes us unique. And to tell you truth, all our cloud customer treat us as a partner, and we have a very good relationship with each one of them. So I would say it's all positive.

Yes. If I were just to bring it home and I'll make my concluding remarks. As you've heard, I mean, for example, Dan's talked about in storage, right? Leading next-generation PCIe interfaces, right? Gen 5, Gen 6 in 5-nanometer. Those are things -- that all goes on to our platform, right? Those are things that the rest of the company can leverage. Same with Raghib's area, right? State-of-the-art CPU cores, right? Neoverse 2, our own fabric and interconnect, that's available to other teams to use as well. We've now got this incredible SerDes and PAM4 DSP capability. You think about connecting our switches to our optical modules, that's an IP set that we have. So I actually -- the way I see it is, each of these groups has their own expertise. We bring it all together, Ashish, and that's really where the uniqueness comes from is being able to go into the customer and then have the way of working for all of the team to really bring in the best of Marvell, every time we go in front of a customer. And we look like one company, one front end. But the access that they get, all the way into the deep, deep, deep sets of technologies and IPs and its circuits, its packaging, its modeling. I mean, its form factors. I mean we have a card business. It -- the list is endless, right, of the capabilities that we can offer. So -- is that the last question? So yes, in closing, thank you so much to everybody that participated today or if you're watching the webcast later. It's been a pleasure to tell our story about the Marvell journey. Personally, while the last 5 years have been incredibly fulfilling and a great run and a great story, I truly believe our best days are ahead of us, and I couldn't be more excited about the future of Marvell. Thank you very much.