Cisco Systems, Inc. (CSCO) Earnings Call Transcript & Summary

Thank you for participating on today's conference call. If you would like to listen to the call in its entirety, you may call 18008392232 . For participants dialing from outside the US, please dial 203

Thanks, Sami, and I appreciate everyone being here with us today. For Cisco, Q3 proved to be an excellent quarter, with our momentum picking up speed and both revenue and earnings per share increasing by double digits while surpassing the upper end of the guidance ranges we had provided. We achieved record revenue of $15.8 billion during Q3, which represents a 12% rise compared with the same period a year earlier. Product revenue climbed 17%, again fueled by strong demand for our AI infrastructure and campus networking solutions. This record top line result, along with operating efficiencies and superb execution from our teams, enabled us to post non-GAAP EPS growth of 10%, which shows how well the initiatives we laid out last quarter are working to offset memory price increases throughout the market. In our view, the confidence our customers and partners have in us has never been more important, and our technology has never been more relevant than it is in the AI era. Consequently, Q3 brought us record high demand. On the whole, total product orders rose 35% year-over-year. If you set aside hyperscaler orders, which increased by triple digits, product orders were 19% higher year-over-year, which demonstrates the broad-based demand we continue to see for our technology around the world. Enterprise product orders increased 18% year-over-year in Q3, with strength seen throughout our entire networking portfolio. Public sector orders rose 27% year-over-year, with double-digit growth in every geography. Product orders from service provider and cloud customers sped up in Q3, rising 105% year-over-year, with 5 of the top hyperscalers each posting triple-digit growth. We likewise saw healthy growth from telco customers in Q3, with orders increasing 9%. Telcos are putting money into Cisco technology as they get their networks ready to handle AI's scale, speed and complexity. Now for some detail on demand from a product standpoint. Networking product orders kept accelerating, rising more than 50% in Q3, powered by triple-digit growth in service provider routing and compute as well as double-digit growth in data center switching, campus switching, wireless, enterprise routing and industrial IoT products. That makes the seventh quarter in a row of double-digit growth for our networking portfolio as a whole. AI infrastructure orders taken from hyperscalers came to $1.9 billion in Q3, versus $600 million in the prior year, with strong growth across our Silicon One systems and market-leading Acacia Optics. The year-to-date figure of $5.3 billion in orders taken from hyperscalers has already surpassed our earlier expectations of $5 billion for FY '26, and we still have a full quarter to go. Because demand is so strong, we now anticipate taking AI infrastructure orders of roughly $9 billion from hyperscalers in FY '26. That is 4.5x our FY '25 total. We expect to recognize about $4 billion in AI infrastructure revenue from hyperscalers in fiscal year '26. Our Acacia business enjoyed its strongest quarter so far, with more than $1 billion in orders in Q3, and it remains on track to grow over 200% year-over-year in fiscal year '26. Acacia leads the coherent pluggable optics market, and we observed strong momentum throughout this business. So far, we've delivered more than 750,000 units of 400-gig coherent pluggable optics, plus over 40,000 of the 800-gig variety, and we think that's well beyond what the next biggest supplier has shipped at either speed. In Q3 we landed 5 new design wins with hyperscalers — 2 in optics, each with a different hyperscaler, and 3 in systems. Those 3 include the first 2 wins for our system powered by Silicon One P200, aimed at major scale across use cases, and a system powered by Silicon One G200 for a scale-out use case. On a separate note, we booked roughly $300 million in AI infrastructure orders during Q3 from Neocloud, Sovereign and Enterprise customers. Order growth has been triple-digit year-over-year in every quarter of fiscal year '26, with about $900 million in orders booked year-to-date, and our pipeline keeps building — approximately $3 billion for our high-performance AI infrastructure portfolio across those customers. Enterprise data center switching orders rose more than 40% year-over-year, and they've now posted double-digit growth in 7 of the past 9 quarters. We think the enterprise AI infrastructure opportunity keeps ramping, since Nexus switch orders tagged for AI deployments climbed almost 50% sequentially in Q3. In campus networking, Q3 brought record orders, up more than 25% year-over-year. Demand for our next-generation switching, routing and wireless portfolio is exceptionally strong, and it keeps ramping faster than prior product launches did; wireless orders hit their highest level ever this quarter, up more than 40% year-over-year. Customers are moving to modern WiFi — orders for WiFi 7 showed strong double-digit sequential growth and made up half of the wireless mix in Q3. Recent research with around 3,500 technology leaders at global enterprises confirms that the urgency to modernize campus and branch networks has increased. With traffic across those networks projected to rise 3x over the next 3 years due to AI, 93% of respondents are speeding up their network modernization plans. These findings back our belief that we're still at the beginning of a multiyear, multibillion-dollar campus refresh opportunity. The strong demand for our technology comes from our ability to deliver AI native capabilities across our products, which includes weaving security into the network fabric and modernizing the operational stack of campus networks. Many of the world's leading companies are investing in Cisco's secure networking solutions because they need high-performance connectivity, automated management and robust security to scale their AI initiatives. Our Cisco Unified Edge solution is also gaining traction, and we've already booked a single enterprise deal for over 1,200 units. Unified Edge combines compute, networking, security, storage and software so AI applications can run at the edge, where data is generated and decisions are made. Our industrial IoT portfolio also had its strongest quarter ever in Q3 and has now grown double digits for 8 consecutive quarters. We expect that demand to continue with manufacturing onshoring to the United States and as agentic and physical AI are expected to drive massive increases in network traffic; now shifting to security. During the third quarter, within our main security offerings, leaving out Splunk, we recorded order increases in the double digits for both newly introduced and updated items, and year-over-year firewall orders rose by a strong double-digit rate. Also, in Q3, more than 1,000 fresh clients bought our recent offerings — Secure Access, XDR, Hypershield, and AI Defense among them — which pushed the cumulative count of net new customers to roughly 5,000 since these launches. Ongoing shrinkage in our earlier-generation lineup still cancels out gains from the new and refreshed set, though not as much as it did during the first six months of the year. Now for Splunk. The movement toward cloud subscriptions and away from on-premise arrangements keeps speeding up, just as we anticipated, and this produces a short-term strain on revenue expansion, consistent with what we described earlier. We anticipate the cloud portion of the business will keep expanding in Q4, and we remain on course to surpass our goal of 1,000 new Splunk customer logos during fiscal year '26. AI is quickening how fast both security defenders and adversaries innovate, and we are advancing rapidly and at scale to assist in establishing an uneven edge for defenders. FY 2025.

Back in March, we unveiled a significant broadening of our Secure AI factory alongside NVIDIA, which hands customers a framework to roll out AI throughout their whole infrastructure—spanning data centers all the way to local sites—so they no longer have to piece together separate, disconnected systems, and security gets baked in right from the beginning. We've likewise rolled out a number of fresh security breakthroughs built to safeguard the full AI life cycle. DefenseClaw is an open source offering that assists customers in safely deploying agents through widely used frameworks like OpenClaw, enforcing guardrails while shielding against malicious behavior and attacks. In order to provide an integrated Zero Trust solution for the agentic workforce, we brought out Zero Trust Access for AI agents, and not long ago we announced our plan to acquire Galileo and Astrix, extending our security and observability platform to cover agentic identity, access management and behavior monitoring; we additionally unveiled new capabilities for the agentic SOC and observability for AI, aimed at helping detect and respond to emerging threats at machine speed and scale. We're collaborating throughout the industry to help defend against AI-enabled threats and to shape next-generation security capabilities. Cisco is a founding member of Project Glasswing and is taking part in private testing of Anthropic's Claude Mythos preview model, built specifically for proactive cybersecurity defense testing. We're likewise involved in OpenAI's Trusted Access for Cyber program. Building on these efforts, earlier this week we announced that Cisco is open sourcing the Foundry Security Spec, a production-grade blueprint for constructing scalable agentic security evaluation systems that use both existing and new AI models. We're handing this blueprint to customers to speed up their ability to capitalize on agentic AI and remain ahead of adversaries. Moving to our innovation and other areas. Cisco IQ, our unified AI-powered delivery engine for Cisco services, has reached general availability, with more than 250 customers already onboarded. Cisco IQ gives customers a real-time benchmark view of Cisco assets and configurations within their environment, helping future-proof it against emerging architectural threats. We also keep accelerating AI advancements internally for our teams. Circuit, our proprietary AI assistant, is now fully embedded into how Cisco operates, with near universal adoption across our employee base and over 8 million total quarterly interactions. Circuit draws on a network of advanced third-party AI models, automatically selecting the best engine for every task or letting users make that choice. As a founding design partner with OpenAI on Codex, our engineers have used it from the beginning. And as of this week, we've made Codex available to our entire product organization so they can build tools and reimagine new products at unprecedented speed. Lastly, we're also proud of our incredible progress in quantum networking. We recently introduced a working research prototype of the Cisco Universal Quantum Switch, designed to route and preserve quantum information between systems at room temperature and over standard telecom fiber. By building this infrastructure now, we're helping accelerate the entire quantum ecosystem that will power the data centers of the future. In order to make sure we seize the major possibilities in silicon, optics, security and AI, a restructuring plan was announced by us today so that resources can be moved around and we can put money into these important growth sectors. These steps are taken from a place of strength and zero in on the technologies that will speed up our expansion, bring unmatched innovation to customers and partners and shape what lies ahead for us. In short, our innovation pipeline is picking up pace and our newest products throughout the portfolio are experiencing some of the quickest uptake in our company's history; this is resulting in wide-ranging record high demand for our technology, which has never been more pertinent to customers than it is in the AI era. This mix, along with the superb execution by our teams, is producing record results and providing value to our shareholders. I'll now hand things over to Mark for further detail on the quarter and our outlook.

Thanks, Chuck. I'm happy to share that Q3 was another solid quarter for us, with revenue and EPS both landing above the top end of our guidance. We set a record for quarterly revenue: $15.8 billion, which is 12% higher than a year ago. Non-GAAP net income came to $4.2 billion, and non-GAAP EPS was $1.06, with both rising 10%. Now let's dig into Q3 revenue a bit more. Product revenue totaled $12.1 billion, an increase of 17%, while services revenue was $3.7 billion, a decline of 1% compared with last year, largely because of when service contracts begin. Networking led the way in product revenue growth, accelerating to 25% versus a year ago, fueled by AI infrastructure and campus refresh; we saw gains across the portfolio, led by double-digit increases in campus switching, data center switching, wireless and service provider routing. Security came in flat, which mirrors the same dynamics we discussed in recent quarters, as growth in new and refreshed offerings keeps getting offset by lower sales of prior generation products and the shift in our Splunk business from on-premise deals toward cloud subscriptions. Collaboration declined 1%, with Webex down, partly offset by devices growth. Turning to our recurring metrics. Total RPO reached $43.5 billion, up 4%. Product RPO rose 6%. Total ARR finished the quarter at $31.2 billion, up 2%, with product ARR increasing 4%. Subscription revenue totaled $7.8 billion, accounting for 49% of Cisco's overall revenue. Software revenue came to $5.7 billion, up 1%. Product orders in Q3 climbed 35% versus the prior year, with broad-based strength. Every geographic segment posted double-digit, accelerating product order growth: Americas rose 35%, EMEA rose 39% and APJC rose 25%. By customer market, triple-digit growth in Service Provider and Cloud led the way. We also saw strength in public sector and enterprise, which grew 27% and 18%, respectively. Total non-GAAP gross margin was 66%, a drop of 260 basis points from a year ago. Non-GAAP product gross margin was 64.3%, down 330 basis points, mostly from negative mix impacts and higher memory costs, partly offset by productivity gains. Non-GAAP services gross margin was 71.6%, up 30 basis points; we remain focused on improving profitability and maintaining financial discipline, with non-GAAP operating margin at 34.2%, which reflects strong execution and operational efficiency. Our non-GAAP tax rate for the quarter was 19%. Moving to the balance sheet. We closed Q3 with $16.6 billion in total cash, cash equivalents and investments. Operating cash flow was $3.8 billion, a decline of 7%, driven by ongoing investments to support rising demand, particularly for AI infrastructure. On capital allocation, we returned $2.9 billion to shareholders in the quarter, made up of $1.7 billion in quarterly cash dividends and $1.3 billion in share repurchases, bringing the year-to-date total to more than $9 billion. There's $9.6 billion left available under our share repurchase program. In summary, we delivered yet another quarter with impressive growth on both the top and bottom lines that came in above what we had anticipated, fueled by solid order increases and a healthy operating margin, which shows how powerful our innovation engine truly is. Our attention stays fixed on strategic investments in innovation so we can seize the considerable growth possibilities we see lying ahead. Those investments will keep being supported by our dedication to careful, disciplined spending management. This potent mix is what keeps driving robust cash flow and our capacity to give substantial value back to shareholders. Additionally, as Chuck said, we are reallocating our resources so we can better grasp the opportunities tied to silicon, optics, security and AI. Under the restructuring plan we announced, we anticipate recording as much as $1 billion in pretax charges, with $450 million being recorded in the Q4 FY '26 and the rest coming during FY '27. Now for guidance. Keep in mind, our Q4 and fiscal year FY '26 guide presumes that current tariffs and exemptions stay as they are through the close of our fiscal 2026. Going forward, you should expect that we will keep prioritizing durable growth, with financial discipline propelling operating leverage and ongoing capital returns. For fiscal Q4, here is our guidance: we expect revenue to fall in the range of $16.7 billion to $16.9 billion; we anticipate non-GAAP gross margin to fall in the range of 65.5% to 66.5%. Non-GAAP operating margin is expected to fall in the range of 34% to 35%. Non-GAAP earnings per share is expected to fall in the range from $1.16 to $1.18; we are assuming a non-GAAP effective tax rate of approximately 19%. Cisco is positioned for its strongest year ever, as shown in our guidance for fiscal year '26, which is as follows: we expect revenue to fall in the range of $62.8 billion to $63 billion. Non-GAAP earnings per share is expected to range from $4.27 to $4.29. Sami, let's now move into the Q&A. FY 2027.

Thanks, Mark. Before we begin the Q&A part of this call, I want to remind analysts to limit themselves to one question plus one follow-up question when they speak. Operator, could we go ahead and bring in the first analyst waiting in the queue?

Nice results — congratulations on those. So, Chuck, to kick things off with my first question: when I look at the guidance for the second half, and in particular what you're projecting for the July quarter, it strongly suggests top-line expansion is speeding up to the low teens, which marks a solid step up compared with the multiyear framework you all have traditionally operated under. Could you address how sustainable this double-digit expansion is? Is it that you're beginning to observe enterprises moving forward with upgrades to their networking architecture, and that's genuinely what's driving it? Or does this represent demand pulled forward? I'd really like to grasp how durable what you're seeing in the second half is; and perhaps I'll pose my follow-up too; it concerns Silicon One. I believe you — you brought up a P200 design win that's set to scale across; that's the first such one you've discussed. Could you take a moment to explain when customers opt for Silicon One rather than other offerings or the merchant offerings available in the market, what value proposition Cisco delivers that makes it so appealing to them? And in terms of the market, how large and how far can this scale for you all?

Thanks, Amit. I'll have Mark begin with the sustainability of growth question, and I'll add my thoughts and then I'll address the Silicon One P200 question.

Yeah, absolutely — thanks, Amit. So obviously we're experiencing plenty of favorable momentum, and we're quite happy with the expansion we're seeing. In Q4, like you said, the top line grew 14.5%, and the bottom line was actually even stronger than that, so those are truly encouraging indicators. Looking ahead to the coming year, we plan to share more specifics and granularity once we wrap up FY '26, and at that point we'll provide a more solid guide. But when it comes to sustainability, there are a few considerations that could be useful to you as you think about FY '27. The first one, I'd say, is essentially — well, it means separating out the AI hyperscale portion of the business from the remainder of the portfolio, so to speak. On the AI hyperscale side, expecting that we'd book no less than $6 billion in revenue during FY '27 seems fair. As I mentioned, a more official guide will come your way in 90 days. For the other parts of the portfolio, assuming growth consistent with our long-term model strikes me as sensible. And keep in mind, the long-term model we initially provided factored in the entire AI hyperscale opportunity too. So the remaining portion of the portfolio could expand, and it's fair to expect that it will do so in line with that guide.

Thanks, Mark. Wait — I still owe an answer to that second question. Regarding Silicon One, the P200 is the clear winner here, so allow me to shed a bit more light on that. In Q3, two separate hyperscalers handed us design wins for the P200 — that's our product, the silicon inside it, deployed for scale-across use cases. Those marked our initial scale-across victories. On top of that, within the opening one or two weeks of the current quarter, we secured a scale-across P200 design win from a third hyperscaler. So the adoption has us genuinely thrilled, and we're quite happy with what the teams are accomplishing on the silicon side. Looking at Silicon One, I'd put it this way: Silicon One is the reason we keep winning. On these calls, I've stated many times across the past couple of years that going forward, lacking silicon means you'll find it hard to stay relevant with the hyperscalers. That, I believe, is exactly what's playing out. So consider the figure we reported and its percentage split — about 50% is systems, meaning Silicon One, and that's a huge differentiator for us. It also — and I'm certain we'll dig into this later in the Q&A — it also grants us far greater command over the supply chain and lets us feel somewhat more assured about our capacity to deliver for our customers; so we're very happy with our current position.

It feels like we've gone back to 1999; wonderful to watch the stock climb and orders rise 35%; that's fantastic. Still, I want to ask about orders, and my angle is the opposite one. Orders outside AI climbed 19%. That leaves out hyperscalers; and right now, for non-AI, conditions aren't really any stronger this quarter than they were the prior quarter. So I'm curious what's behind that? And does it worry you that enterprises — customers outside AI — might be pulling purchases forward due to supply limits, due to trouble receiving products on schedule. I'm just asking whether this sort of growth can be sustained.

Thanks, Tal; so what are we observing today that we believe is fueling this pickup? First off, we're seeing AI emphasis within the enterprise keep broadening. Customers are now getting ready for inferencing and agentic applications; and in such situations, the network matters enormously and shifting the bits around with low latency matters enormously and customers are waking up to the fact that they must modernize; that's the reason our enterprise data center switching business came in with orders up more than 40% for the quarter. And that's purely enterprise build-out; so we see that picking up speed. Second, driving this is modernization, but we also think customers are getting ready for higher levels of cybersecurity threats. We'll say a little about Mythos shortly. But I believe some customers are beginning to work through the consequences and what they must do to be ready for that. And clearly, we keep seeing design wins in the hyperscale space and ongoing uptake of both our silicon-based products and our optics; so that's roughly how we view what's taking place. And Mark will go over some of the figures. I also don't believe you'll find that our lead times on the traditional networking side, Mark, keep me honest, but they aren't extreme at the moment. So would you like to discuss the figures a bit, Mark?

Okay, so this is something we've gotten a lot of questions on, and it was clearly the main focus of the notes that came out ahead of the earnings release — I'm talking about pull aheads, or pull forwards. To kick things off, I'd say it's reasonable to think that some level of pull ahead did happen in Q3. Figuring out the exact magnitude is genuinely tough, but our sense is that it amounted to a very small number in Q3, and I'll give you three pieces of evidence for that view. The first one — and you raised this point — is that if you look at the underlying growth rate excluding webscale, our order growth in Q2 came in at 10%. Turning to our third-quarter order growth rate, excluding webscale, that figure reached 19%. That means you witnessed an acceleration of approximately 9 points. When we work out the contribution from the price increases we put through, those account for around half of the pickup — roughly 4 to 5 points of additional growth, with the same units but at elevated prices. As a second point, we routinely look at pipeline pull forwards, meaning orders pulled from later quarters. Naturally, some degree of that always takes place. When we examine the deals from Q4 that got moved up and closed during Q3, we honestly saw no real distinction or extra pipeline shifted earlier from Q4 into Q3 — pardon me, I mean from Q4 into Q3 — relative to what you'd find looking at that same stretch of time one year back. The next point I want to share concerns the Q4 pipeline on its own. That pipeline looks quite robust, with strong growth compared to last year; and the situation was that across weeks 1 through 13 of Q3 — meaning we're discussing the Q4 pipeline here — there was zero deterioration in that Q4 pipeline from when Q3 started until it finished. It even increased over the course of that quarter. So when you add it all up, once again, pinning down the exact figure is tough. Some portion probably exists, but we think it's on the smaller side.

Normally I stick to a personal guideline of offering praise to leadership only when shareholders are seeing gains, and in this case it's deserved; so I appreciate the question. I'd like to dig into what Mark said regarding the portion that isn't AI expanding at a rate matching the model for the coming year, 2% to 5%, if I recall correctly from what you stated during your Analyst Day. Perhaps I've got the figure wrong, perhaps mine is too conservative. But my thinking would be that for the non-AI part, the price hikes should — despite facing a difficult comparison by the 3Q — the price hikes should essentially push you beyond that level. And so I was hoping you could break that down a bit further, whether the non-AI piece, when you reference the long-term model, really is 2% to 5%, and wouldn't the price hikes lift you past that considering all that we're observing and these robust trends.

Ben, this is Mark. So my view is that regarding the long-term model, the figure was in fact 4% to 6% overall. So the 2% to 5%, as I see it, applied to the networking segment by itself. We actually experienced a considerably greater expansion rate on security and observability, for example. And so essentially what we're conveying is, to begin with, we'll provide a far more thorough and precise outlook once we move through FY '26. So 90 days from now, we'll discuss it with much greater specificity. But I was merely noting that, that 4% to 6%, excluding the AI hyperscale opportunity, you could anticipate that we would expand at a rate consistent with that. And once more, as you pointed out, we'll face some more challenging comparisons as we head into the following year. The price hikes ought to benefit us too, and we'll say more a quarter from now.

I’ll put both of my questions out there together. I suppose the first one goes back to the supply chain. A competitor of yours said not long ago that conditions were becoming even more constrained, and they even mentioned the possibility of decommits. So I’m wondering whether you could walk through your view of what’s happening across the broader supply chain, perhaps not only in memory, including what you’re doing to secure wafer allocation for Silicon One, or anything else you’re able to say about that? Then my brief follow-up: your AI order intake into fiscal 4Q implies a very large step-up, from $1.9 billion — if I’ve done the math correctly, that’s close to $3.7 billion. What I’d like to know is what’s behind that? Is it breadth across wins? Are you beginning to see scale-up activity building in your pipeline? Could you break down that pretty striking increase for the coming quarter?

Sure, I’ll say something briefly, and then, Mark, you can address the supply chain work we’ve undertaken. I want to remind everyone that this is among the major benefits of Silicon One. That’s one of the key advantages. A much larger portion of our supply chain sits under our control, and we don’t face as many dependencies as some others might. But Mark, would you like to cover some of the — I think you ought to discuss a few of the steps we’ve also taken, and then how we’re feeling, along with the fact that we really haven’t encountered any decommits whatsoever. Would you like to walk through that...

Absolutely, Chuck. So, yeah, I'd agree with what Chuck mentioned — the reality that we handle our own chip design definitely provides us with tighter oversight across the whole chain. I mean, the simple fact that we're overseeing wafers, substrates, assembly, and test directly gives us significantly more command over the supply chain, so to speak. On the silicon side, we've locked in our supply for the next 8 months, through calendar year '26 anyway. After that, the usual talks are live and in progress for calendar '27. Turning to memory, several efforts have been launched. We've got more than 20 programs up and running that aim to cut memory usage throughout the portfolio — one illustration being in wireless, where you'll find products becoming orderable in Q4 that need 50% less memory; that's a major win. Another point is that we're putting money into fresh capacity. You've likely heard about the strategic investment we announced — plus a 3-year supply agreement with them. That will be a real boost for us. The move we're making in several locations, shifting from DDR4 over to DDR5, along with those conversion efforts currently in progress, is also quite favorable. Broadly speaking, with inventory and advanced purchase commitments, our financial strength lets us push hard on that front. Taken together, that's a $6.7 billion rise within just the last 90 days, which works out to 48%. Looking at it year-over-year, once again, inventory and advanced purchase commitments have climbed $11.6 billion year-over-year; so overall, whether we're talking silicon, substrates, memory, photonics, PCBs, or power, we're locking in long-term agreements wherever we can. We're collaborating with our sub-tier suppliers, and we're stockpiling strategic inventory; that's a major point of differentiation. And let me just echo Chuck's comment — we saw zero decommits during the quarter.

Absolutely, and regarding your next question about the forces behind the AI orders — those large AI orders. I'd say most of what we'll witness in Q4. To start, I want to point out that we've described this business as nonlinear. When you look at the chart included alongside the prepared remarks, it clearly shows Q3 came in at 1.9, slightly below Q2, followed by an acceleration in Q4; so many timing dependencies exist here, which is a reminder that things stay nonlinear going forward. These are Silicon One wins, after all. Optics accounted for half of the wins. Our optics business truly sped up. Acacia is performing exceptionally well; we had — much of it is scale out, mainly scale out. We've just received word on the scale across. So in fact, scale across isn't reflected in the numbers at this point. Out of those 5 design wins we mentioned, we'd anticipate seeing some initial orders during Q4 and — though not at scale until we get into fiscal year '27. So I believe it stems from a great deal of relationship-building the teams have done across the years. They appreciate our silicon. They appreciate the supply, and they appreciate our optics. And that's — that's essentially it. So we simply plan to keep working very hard and provide whatever they require.

Perhaps I could follow up on that previous question. When comparing the AI revenue figures you've provided for fiscal '26 and fiscal '27, the AI orders clearly don't stretch the longest in terms of duration. Still, I was hoping to understand how much longer those orders are running, given worries about the supply chain that might not affect you as directly but certainly do matter further along it. Then, as a second thing, is there any way to break down the gross margin pressure on product between memory and product mix?

Mark, would you like to handle those?

Yes, thank you, Meta. I want to address this straight away. As far as how long these things take, we aren't really observing anything out of the ordinary. Our position has always been that these orders don't arrive on a predictable schedule, customers make their plans far ahead of time, and none of that has changed. Turning to gross margin, the teams deserve real credit. Our Q3 gross margin came in at 66%, which matches what we anticipated, aligns with our plan, and sits precisely at the midpoint of the guidance we gave. We're confident gross margin has found its footing. That shows up in the Q4 outlook too, where the midpoint again lands at 66%. As I noted, there are several items related to the more than 20 programs we're running focused on memory utilization. The teams are executing really well on those. Then there's the shift from DDR4 to DDR5. On top of that, several actions we described three months ago are now in motion — the price increases you brought up, certainly, plus reworking our terms and conditions, and truly stepping up to use our balance sheet strength through advance purchase commitments. You can see those commitments climb by roughly $6 billion over the past 90 days. Beyond gross margin, I'd add that our attention is on generating operating leverage, and our earnings are expanding more quickly than revenue. That's visible in the Q4 outlook, and it's also visible in the complete FY '26 picture and the guidance we provided. A crucial route to operating leverage involves sharp attention to operating margins themselves; Q3 gave us record operating margin dollars. Our goal is to hit 34% operating — that is, operating income as a share of revenue. Once more, that has shown up each quarter during FY '26. It appears in the annual view too. We can also make certain trade-offs; over the longer horizon, should gross margin come under pressure, I believe there are many levers available to us. One illustration from Q3: even as gross margin fell 2.6%, operating expenses as a share of revenue dropped by over 2% as well. Twelve months earlier it stood at 34.1%, and this quarter it was 31.9%. So I expect we'll enter some businesses with differing scale. That means we have plenty of options to safeguard that 34% operating income.