Akamai Technologies, Inc. (AKAM) Earnings Call Transcript & Summary
November 6, 2025
What were the key takeaways from Akamai Technologies, Inc.'s November 6, 2025 earnings call?
In the third quarter of fiscal year 2025, Akamai Technologies, Inc. reported revenue of $1.055 billion, reflecting a 5% year-over-year increase, and non-GAAP earnings per share (EPS) of $1.86, up 17% year-over-year. The strong performance was driven by significant growth in Cloud Infrastructure Services (CIS), which saw a 39% increase in revenue, and robust demand for high-growth security products. Management raised guidance for CIS annual recurring revenue (ARR) growth to a range of 40% to 45% for year-end 2025, signaling strong momentum in this segment and a positive outlook for the company's future growth prospects.
What topics did Akamai Technologies, Inc. cover?
- Cloud Infrastructure Services Growth: CIS revenue reached $81 million, up 39% year-over-year, accelerating from 30% growth in Q2. Management noted, 'We continue to expect CIS ARR year-over-year growth in the range of 40% to 45% in constant currency at year-end.'
- Strong Security Product Demand: Revenue from high-growth security products grew 35% year-over-year, driven by strong demand for API security solutions. Management stated, 'We now expect to exit 2025 with a run rate of approximately $100 million for that product line.'
- Delivery Revenue Stabilization: Delivery revenue was $306 million, down 4% year-over-year, but management indicated this was slightly better than expected, marking continued stabilization. They noted, 'This result was slightly better than expected, marking another quarter of improved trends in our delivery business.'
- AI Inference Cloud Launch: The launch of Akamai Inference Cloud is positioned as a key growth driver, with management highlighting strong customer interest. CEO Tom Leighton remarked, 'We're in the first inning of a very exciting game,' indicating significant future potential.
- CapEx and Share Buyback Strategy: CapEx for Q3 was $224 million, representing 21% of revenue, as the company continues to invest in CIS. Management confirmed, 'We spent $800 million to buy back approximately 10 million shares,' marking the largest annual buyback in history.
What were Akamai Technologies, Inc.'s November 6, 2025 results?
- Revenue: $1.055 billion (vs $1.02 billion est, +5% YoY)
- Non-GAAP EPS: $1.86 (beat by $0.20)
- CIS Revenue: $81 million (up 39% YoY)
- Security Revenue: $568 million (up 10% YoY)
- Delivery Revenue: $306 million (down 4% YoY)
- Operating Margin: 31% (improved YoY)
Akamai's strong Q3 results and positive guidance signal robust growth potential, particularly in Cloud Infrastructure Services and security products. The launch of Akamai Inference Cloud presents a significant catalyst for future growth. Investors should monitor the performance of the delivery segment and the overall demand environment as potential risks moving forward.
Earnings Call Speaker Segments
Mark Stoutenberg
executiveThanks for spending time with us this afternoon for the call covering Akamai's results from the third quarter of 2025. Two people are presenting: Tom Leighton, who serves as our Chief Executive Officer, and Ed McGowan, our Chief Financial Officer. Before we go further, a reminder that what we say today contains forward-looking statements, among them guidance on revenue and earnings. Those statements rest on expectations and assumptions we hold right now, and they carry risks and uncertainties along with several factors that might make actual outcomes differ materially from what we express or imply. Such factors encompass, though aren't limited to, effects arising from macroeconomic trends, the integration of any acquisition, geopolitical developments and other risk factors identified in our filings with the SEC. Everything shared during this session reflects where the company stands as of November 6, 2025, and we're under no duty to revise any forward-looking remarks later on. Also, please note that some of the figures we'll discuss today are non-GAAP measures. You can find a full reconciliation between GAAP and non-GAAP numbers in the financial materials posted within the Investor Relations area of akamai.com. With that said, I'll pass things over now to Dr. Tom Leighton, our CEO.
F. Leighton
executiveMark, thanks. I'm glad to report that Akamai had a strong third quarter, with results surpassing what we expected for revenue, margin, and earnings per share. Revenue climbed to $1.055 billion — a 5% increase from a year ago on a reported basis, and 4% higher in constant currency. Non-GAAP operating margins landed at 31%, which marks an improvement, and non-GAAP earnings per share hit $1.86, up 17% versus last year both as reported and in constant currency. Throughout our portfolio, performance was solid: Cloud Infrastructure Services, or CIS, gained momentum, demand stayed strong for our high-growth security products, and delivery revenue kept stabilizing. One thing we're especially pleased to see is that customers and industry analysts are increasingly recognizing the power of our distributed platform and our distinctive approach. If you're an investor who isn't as familiar with the way we've reshaped Akamai's business model — taking it from being the CDN pioneer to becoming a leader in cloud security and distributed cloud computing — I'd suggest taking a look at a new IDC report about Akamai called "Akamai: Navigating the Cloud Frontier — A Transformation from CDN to Distributed Cloud Provider." In it, you'll find an unbiased third-party view of how Akamai has developed its strategy and its main points of differentiation in security and distributed cloud computing for AI inferencing at the edge. You can find the report on our website. To illustrate just how much Akamai's distinctive distributed compute approach can deliver, all three of the biggest cloud providers in the U.S. are now customers of Akamai Cloud Infrastructure Services. What's more, during Q3, one of those three inked a multiyear renewal that was expanded, cementing Akamai's standing as their go-to provider for premier distributed cloud computing. This particular customer holds a dominant place at the Internet's core, and relies on our managed container service, which is widely distributed, to bring their business logic nearer to end users so performance improves. Cloud infrastructure services revenue during Q3 reached $81 million, which is a 39% increase year-over-year both as reported and in constant currency — a faster pace than the 30% growth we saw in Q2. During the third quarter, we closed a number of fresh agreements and extensions tied to our Cloud Infrastructure Services. Among those clients were a prominent South Korean maker of appliances and consumer electronics with worldwide reach; a financial services firm operating across multiple countries that is based in Singapore; a top U.S. analytics software developer; a supply chain planning software seller headquartered in the United States; a cybersecurity company from Europe; a significant U.S. airline; a prominent American video game business; a leading Indian media and entertainment organization; and one of the biggest media firms globally. In addition, a Japanese multinational gaming enterprise signed on for our Cloud Infrastructure Services, which formed part of a broader two-year, $37 million renewal covering a range of Akamai products and services. During last week's AI Industry Conference, NVIDIA GTC, I joined our team as Akamai made a significant move toward what lies ahead by introducing Akamai Inference Cloud — our platform built to meet rising demand for scaling AI Inference across the Internet. As AI grows, the Internet's architecture is experiencing a basic transformation. What we are building today with the Internet is being shaped by AI, where intelligent systems powered by AI Inference strengthen and augment human intelligence. In the same way we deliver the web, Akamai is prepared to deliver inference by positioning it physically close to those who use it. This enables faster performance and worldwide scale, which in turn supports smart applications everywhere on the planet. Back when the web was just emerging, the demand for speed and scale is exactly what set Akamai's creation in motion. We played a part in putting an end to the so-called World Wide Wait, letting the Internet grow so it could deliver real-time services to billions of individuals worldwide. After that, we rolled out web security as a cloud service, which let the web be used safely for all sorts of vital applications, including banking and commerce. With AI inference right now, we observe that same requirement for speed, scale and security unfolding once more. Through uniting highly scaled GPU and compute capacity with Akamai's unmatched worldwide reach and edge security, Akamai Inference Cloud lets intelligence operate immediately, safely and precisely wherever it's required, directly beside the user, agent or device. That's how Akamai is able to fuel the upcoming wave of AI applications — conversational, personalized and Agentic. Each one built to scale in real time to satisfy unmatched demand. Looking at AI investment cycles, we find the market sitting at a turning point. Up to this point, the AI narrative has mostly centered on training, the first creation of AI models out of huge quantities of underlying data. To train foundation models, AI pioneers have depended on hyperscale clouds and their centralized data centers, which hold tremendous concentrations of compute, power and capital; our view is that AI Inference, meaning the execution of queries against a trained model, represents the new frontier — one demanding purpose-built infrastructure so that distributed low-latency, globally scalable inference at the edge becomes possible, with response times measured in a few tens of milliseconds. As AI systems get adopted at scale, we anticipate that the growth of Inference will generate massive demand for this new intelligent layer of the Internet. And we aren't alone in anticipating it. Fortune Business Insights, in a report covering the expanding worldwide AI Inference market, noted that Edge Inference holds the leading position in that market, driven by escalating need for AI processing that happens close to data sources with real-time, low-latency performance, and it is expected to achieve the highest CAGR among all AI inference models. Jensen Huang, NVIDIA's Founder and CEO, also stated, regarding the Akamai Inference Cloud launch at GTC, that "Inference has become the most compute-intensive phase of AI, demanding real-time reasoning at planetary scale. NVIDIA and Akamai have joined forces to bring Inference nearer to people around the world, providing generative AI that is quicker and scales better, while opening the door to a new wave of smart applications. By combining Akamai's worldwide distributed setup and know-how with NVIDIA's Blackwell AI infrastructure, Akamai Inference Cloud delivers the processing power needed to reveal AI's full capabilities. As of now, customers can reach this offering in 17 cities across the globe, and more points of presence are on the way as demand from customers keeps growing. Monks — a European company focused on digital-first marketing technology services and consulting — is among our first clients, and they remarked that using Akamai Inference Cloud will speed up how quickly they roll out important features, such as recognizing players and plays and getting tactical insights to coaches in real time during a game. Distributing advanced GPUs out to the edge is the only way that becomes achievable, and in their view it will reshape how they handle sports broadcasting along with immersive fan experiences. Harmonic, another client, provides technology for distributing video content across television and the Internet, and they stated that Akamai Inference Cloud lets them operate larger parameter, more capable models on local infrastructure, which widens the set of functions they can offer affordably inside a single compute instance, resulting in quick response times, refined personalization and richer video content. Here at Akamai, our view is that the technology ecosystem powering the AI revolution is going to need several suppliers of AI infrastructure — the hyperscalers, NVIDIA, and Akamai too, given our distinctive distributed capabilities and unmatched edge expertise. We couldn't be more enthusiastic about the future. Naturally, the edge is likewise where Akamai delivers our security offerings.
F. Leighton
executiveWhen customers discuss their AI inferencing strategies with us, they mention a valuable synergy they perceive between Akamai's security and delivery offerings and our cloud computing strengths, along with their confidence in Akamai to enhance Internet speed, reliability, and security for their operations. During Q3, Akamai's security growth was fueled by robust demand for our leading segmentation product and by swiftly rising adoption among customers of our API security offerings. Together, these fast-expanding security offerings boosted revenue by 35% year-over-year as reported and 34% in constant currency. Our segmentation victories in Q3 featured a $3 million expansion deal to provide one of North America's biggest health care technology firms with the visibility and control they previously lacked. A $1 million agreement with a European insurance group that is also a brand new Akamai customer, a multiyear deal with a major insurance company in Korea; plus an expansion agreement with a large bank in Mexico to broaden their initial rollout across their Latin American operations. In Q3, we likewise saw escalating interest in our top-tier API security solution. As organizations move toward an API-first approach and broaden their use of AI applications that depend fundamentally on APIs, they adopt tools that let them uncover and track deployed APIs and handle risk. Meet tougher data protection rules, particularly in Europe, and address public disclosures of API-linked breaches that have heightened awareness and elevated the consequences for financial and reputational damage. Akamai API security victories in Q3 featured a $7 million deal for API security with one of Europe's most significant banks. As part of a $31 million multiyear commitment for security and compute. The uptake of API security within a $20 million expansion agreement with one of the world's biggest airlines. An expansion of API security as part of a $42 million deal spanning the full range of our portfolio with one of the world's largest software firms. A $2.6 million agreement with one of Asia's biggest life insurance groups. The replacement of a rival at a FinTech payments company in Brazil. And we additionally secured 7-figure deals for API security with 2 of the 10 biggest financial institutions in the U.S. and one of the largest banks in Canada; we're also happy to report that for the sixth consecutive year, Akamai has earned Customers' Choice in Gartner's most recent Voice of the Customer report for cloud web application and API protection. Akamai was likewise acknowledged as Customers' Choice in Gartner's Voice of the Customer report on online fraud detection. Before handing the call to Ed, I want to convey my appreciation to our team members for their outstanding efforts in Q3, and for making Akamai an excellent workplace. In acknowledgment of your contributions, Glassdoor ranked Akamai among their top 50 best led companies of 2025. At this point, I'll hand things over to Ed, who will go into further detail about what we saw in Q3, along with what we expect for Q4 and the full year. Ed?
Edward McGowan
executiveThanks, Tom. I'm glad to report that we had a strong third quarter. Total revenue was $1.055 billion. On a reported basis, that's up 5% from a year ago, and 4% in constant currency. We also delivered another quarter of very strong bottom line results. Non-GAAP EPS was $1.86 per share, which is a 17% year-over-year increase, both as reported and in constant currency. That strong EPS number was driven by revenue that came in above our expectations, along with solid execution across the organization. Now let's turn to revenue. Compute revenue came in at $180 million. This includes the Cloud Infrastructure Service, or CIS, solutions that are expanding quickly and that Tom mentioned earlier, as well as our other cloud applications, or OCA. On a reported basis, that represents an 8% increase compared with the same period last year, and 7% when currency effects are excluded. Keep in mind that Q3 2024 included a $7 million one-time gain from deferred revenue being released after a long-term legacy compute contract came to an end. Those dollars came from our other cloud application products, and they added 5 percentage points to the year-over-year growth rate for total compute revenue. Total compute revenue got a lift from continued momentum in CIS; during Q3 2025, CIS revenue came in at $81 million, with growth accelerating to 39% year-over-year on both a reported and constant currency basis, a solid improvement over the roughly 30% growth seen in the previous quarter. So with all that in mind, our outlook for CIS ARR growth at year-end is still 40% to 45% higher than last year on a constant currency basis. On the security side, revenue reached $568 million, which is a 10% increase compared with a year ago as reported, or 9% in constant currency. Our high-growth security offerings — meaning API security and Zero Trust Enterprise Security — brought in $77 million, up 35% year-over-year, and 34% in constant currency. Given how well our API security business keeps performing, we now anticipate finishing 2025 at a run rate of roughly $100 million for that product line, on both an as-reported and constant currency basis. To wrap up, for 2025 we still anticipate that combined ARR from our high-growth security offerings will grow 30% to 35% compared with the prior year on a constant-currency basis. Now on to delivery. That segment posted $306 million in revenue, a 4% decline versus a year earlier both as reported and in constant currency. It came in a bit ahead of what we had anticipated, and it represents yet another quarter of improving momentum in our delivery operations. International revenue reached $525 million, rising 9% from a year ago, or 8% in constant currency, and accounted for 50% of our total revenue in Q3. Currency movements helped revenue by $4 million sequentially and by $8 million on a year-over-year basis. Let's shift now to profitability. For the third quarter, non-GAAP net income came in at $269 million, which works out to $1.86 per diluted share. That's a 17% increase compared with the same period last year — impressive growth that holds both as reported and in constant currency — and it beat the top end of our guidance range by $0.20. Turning to capital expenditures, Q3 CapEx totaled $224 million, representing 21% of revenue, reflecting our ongoing investments in the CIS business, which is growing quickly. Now let's look at cash and how we approach capital allocation. As of September 30, we held approximately $1.8 billion in cash, cash equivalents, and marketable securities. During Q3, we did not repurchase any shares. As a reminder, through the year so far, we've spent $800 million buying back roughly 10 million shares — the largest annual buyback in our history. When it comes to how we deploy capital, our plans haven't changed: keep repurchasing shares over time, offset the dilution caused by employee equity programs, and stay opportunistic on both M&A and buybacks when market and business conditions justify it. Before I walk through our Q4 and full year 2025 guidance, I'd like to mention a few housekeeping points. First, just like last year, seasonality is a major factor shaping how we perform financially in Q4. Normally, we experience above-average traffic from our large media customers, along with a seasonal uptick in online retail activity from our e-commerce customers. That said, predicting either one is tough. Second, on Q4 operating expenses — as is customary for our business, we anticipate they'll come in above Q3. The primary reason is the seasonal spike in sales commissions when our top-performing reps hit their annual quota accelerators. Lastly, on July 4, 2025, the One Big Beautiful Bill Act became law, bringing in notable provisions like permanently extending certain expiring Tax Cuts and Jobs Act provisions, modifications to the international tax framework, and it reinstated some business tax benefits. That said, this new legislation hasn't materially affected our tax rate in 2025. With all that in mind, I'll turn to our Q4 guidance. Looking ahead to the fourth quarter, our revenue projection falls somewhere in the range of $1.065 billion to $1.085 billion. That represents growth of 4% to 6% as reported, or 3% to 5% when measured in constant currency, compared with Q4 2024. Based on spot rates right now, we expect foreign exchange movements to have a negative $5 million effect on Q4 versus Q3 levels, while the year-over-year impact would be a positive $11 million. Turning to the full year, our guidance at current spot rates assumes foreign exchange will contribute a positive $13 million to revenue in 2025 on a year-over-year basis. Assuming those revenue levels, we anticipate cash gross margins of approximately 72% to 73%. For the fourth quarter, our non-GAAP operating expenses should come in somewhere between $322 million and $331 million. On the EBITDA side, we're looking at a margin of about 42% to 43%. Non-GAAP depreciation is anticipated to fall in the $143 million to $145 million range, and non-GAAP operating margin for Q4 should be around 28% to 30%. When it comes to CapEx, the figure we're expecting is somewhere in the neighborhood of $171 million to $181 million, which represents about 16% of the total revenue we're projecting. Based on how we see revenue and costs shaping up, non-GAAP EPS for the fourth quarter should land between $1.65 and $1.85. This non-GAAP outlook factors in taxes of $57 million to $60 million, reflecting an estimated quarterly non-GAAP tax rate of roughly 18% to 19%. That number assumes about 147 million shares on a fully diluted basis. Our complete 2025 full-year guidance is available in today's press release, though I'd like to walk you through the highlights right now. For the entire year, we expect constant-currency total revenue growth between 4% and 5%. Non-GAAP operating margin should land near 29% to 30%, with earnings per share in the $6.93 to $7.13 range. Wrapping up, we continue to be very pleased with the results from our Cloud Infrastructure Service and our security offerings in high-growth areas. We're quite excited about what the Akamai Inference Cloud can do, since it takes AI to the edge. That said, I'll bring my remarks to a close. Tom and I would be glad to answer your questions. Operator?
Mike Cikos
analystNice work on the strong quarter and the execution you just described. I want to circle back to what you said about the segments real quick, just to confirm I've got the guidance on each piece right. On security and compute, at this point are we repeating the growth figures you mentioned three months ago? My recollection is you were aiming for roughly 15% in compute and about 10% in security. Could you give us any last specifics on those? And I have one more question after that.
Edward McGowan
executiveMike, it's Ed here. Sure — for security, we're guiding to roughly 10%. We don't break out guidance by quarter. So on a full-year basis, we clearly reaffirmed our ARR guidance; nothing there has moved. And like we've said previously, on compute it'll land perhaps slightly below 15% for the year, since a few of the larger contracts came on a bit later in the year than we'd anticipated, though momentum in CIS is certainly building.
Mike Cikos
analystExcellent; and before I move on, one additional item for Tom. When we consider '26, particularly in light of how '25 has unfolded so far, is there anything you can share—any hints or boundaries?
Edward McGowan
executiveNo; guidance for '26 will come during the February call. What I can tell you is that the momentum we're observing in CIS, particularly regarding the new AI Inference Cloud, has generated an enormous amount of activity and demand; as a result, there's a strong possibility we might speed up growth in our CIS business next year.
Mike Cikos
analystThat's great. And that leads perfectly into what I was going to ask next. Tom, I realize you devoted a large portion of your prepared remarks to Akamai Inference Cloud, and that has to be deliberate, correct? You mentioned a few early customers, including Harmonic. You brought up the 17 locations and the fact that this is offered worldwide. Could you just give us a sense of where things stand on that inference curve? We've been living in the frontier pioneer model era for a couple of years now. It seems like momentum is building at the inference layer, but what does the on-the-ground picture look like from where you sit today?
F. Leighton
executiveYes, we're in the opening inning of a game that's going to be very exciting. There are simply so many customers eager to embrace AI applications and perform inference for — applications that must be quick for real users. Take a commerce company as an illustration: they wish to surface the right items, personalized images to the shopper, and then when the shopper shows interest in something, they truly want to display not just the item, but the shopper wearing it, and not merely in a still image, but in a video; and within that video, they want to place the shopper in a setting and surroundings that feel appropriate for that shopper. We have customers with AI-powered toys. Customers with — who are pursuing robotics and require extremely fast control and response. So they need to run the compute, the inference near the end user at the edge. Media Workflow, people will encounter various versions of games, various versions of movies. Everything will be personalized. And this calls for something such as the NVIDIA Blackwell 6000, which is extremely powerful, yet also ideally suited to operate at our edge. That's precisely why partnering with NVIDIA is so sensible. So we're in the opening inning. We only began rolling out the service over the past couple of weeks, but we're witnessing a very high level of interest across our customer base. This is precisely what they're looking for.
Lawrence Vensko
analystLawrence Vensko here, filling in for John DiFucci. You've laid out what you're doing on the go-to-market front — bringing in hunters and seasoned specialists to back new business sales along with security and compute. Could you take us through your thinking on adding more reps over the rest of this year? And what should we expect regarding how long it takes those reps to ramp up before they're at full productivity?
F. Leighton
executiveYes. Our go-to-market transformation has moved forward considerably, and it will keep going into the early portion of next year. By the time Q2 arrives, I'd say much of that transformation will be finished. Hiring will continue from now through at least the first half of next year, building greater sales capacity for hunting in security and, clearly, in compute as well. So that effort carries on, though the transformation piece will be mostly complete around the start of next year.
Rishi Jaluria
analystFantastic. That's good to hear — clearly lots of momentum, particularly when it comes to compute. I might have a couple of questions. To begin, Tom, you mentioned certain agreements you're closing with the hyperscalers and the ways they're relying on you for additional edge inferencing. Yet at the same time, as all of us understand regarding the hyperscalers, at present everybody is dealing with limits on capacity, and everyone is hunting for whatever resources are available anywhere, given how we're trying to push AI forward in both training and the inferencing/reasoning space. So perhaps you can help me see why you're confident that once we move past this period of constrained capacity — no matter how long that takes — Akamai will remain in a strong position to gain from it. And that it isn't something a few of the hyperscalers or partners would consider taking in-house once the opportunity becomes real, particularly with edge inferencing and as additional capacity gets constructed? Then I have a brief follow-up.
F. Leighton
executiveYes. Akamai has a platform unlike any other. No one else is comparable to us. We have more than 4,000 points of presence. We're already present in over 700 cities. That is distinctive. And we already provide our Edge Worker solution, our function of the service, at every location. As a matter of fact, one hyperscaler relies on us for ad market, so their ad logic sits much nearer to users, while another hyperscaler relies on us for API orchestration, which again has to be quick and near users. We discussed our managed container service. A third hyperscaler relies on our managed container service for their media workflow, once again since it's nearer to users; so this isn't a matter of capacity limits; it isn't that the hyperscalers have exhausted their capacity. Surely, that isn't true. It's because our platform is distinct and we're able to make their logic, their compute logic, run more quickly for users since it's nearer. And clearly, we go up against their cloud businesses to secure that business, but that's thanks to our superior performance. And the positive part is we don't need to go deploy 4,000 new reasons or begin new deployments across 700 cities, since we already possess this distributed platform. We're expanding several of the regions we've discussed with the new Blackwell 6000, and that's quite thrilling. And we'll keep doing so as we grow that business.
Rishi Jaluria
analystYes, got it. That makes sense. So maybe now I'll move on — thinking about the API Security side of things. Clearly, you've had success there; beyond that, when we consider how applications are evolving, correct, moving toward agents and multi-agentic systems; and clearly, things such as MCP and A2A are emerging as a means of pulling all of that together. Could you take what's happening in API security right now and begin expanding it into these new Agentic protocols?
F. Leighton
executiveDefinitely, and that's something we're already engaged in. Take, for instance, API Security and spotting the APIs, spotting the new AI agents and apps — the shadow AI that companies are running without realizing it. And then, naturally, safeguarding that through our newly launched AI firewall service. That's among the reasons we're extremely enthusiastic about our API security business. So it's not only about safeguarding legacy APIs, but also the new Agentic web and safeguarding — plus identifying and safeguarding all those emerging inferencing applications.
James Fish
analystThat's fascinating on the compute front, but how are we supposed to view the CapEx needed for inferencing in this case, and what's the current nature of your relationship with NVIDIA? Has it changed at all across roughly the past 12 months? And on that same topic for you, how are we supposed to think about compute gross margins going forward across the coming year to 2?
Edward McGowan
executiveJim, those are good questions. An initial deployment has been completed alongside NVIDIA. We're present in roughly 17 cities, and we have that operational; that fell within our Q4 CapEx; it's going to be guided by demand. What we observe at the outset is certain customers approaching us with fairly substantial requests. Consequently, revenue will trail the CapEx quite tightly, and demand will be what guides it. Earlier we provided you a rough metric of approximately $1 of CapEx corresponding to $1 of revenue. Here it's roughly the same. Depending on utilization, it might be somewhat better. When it comes to margins, they'll be quite comparable. Naturally, I'd anticipate that during a deployment, suppose we receive orders worth $50 million, $100 million, or whatever the figure might be, and you're spending a bit more CapEx for things further ahead — pipeline-related items. At the start it'll be somewhat inefficient; yet after reaching scale, I'd anticipate gross margins comparable to those we see in compute, and possibly somewhat superior simply because of scarcity in the market. So slightly better margins are a possibility. Beyond that, we also anticipate achieving fairly solid operating leverage and scale from it; thus as we grow — as we scale up, I anticipate that we'll get — assist the company's overall gross margin once we reach scale.
James Fish
analystUnderstood. Turning back to the go-to-market aspect. During this quarter, there appeared to be numerous mentions of, let's say, cross-selling or bundling spanning all 3 segments. Can we get some clarity on the incentives being offered to your sales force? Or what proportion of the business consists of what you might term ELAs, while others refer to it as pool of funds. At this point, what share of customers utilize all 3 segments?
Edward McGowan
executiveYes. That's a fair question. One thing we implemented, and what we're observing out in the field this time around this year, is that we provided incentives for longer-term deals. Consequently, we're certainly witnessing a lengthening of our sales cycle — actually not sales cycle, but our average customer duration, which is positive; so longer-term deals, bigger deals, and so on are showing up. That will be evident in the RPO metric scheduled to be released tomorrow in the queue. So that's among the key factors. We're also noticing bigger deals, though not necessarily pool of funds or ELAs. A handful of those exist, but they aren't necessarily the main factors. Adoption of compute is clearly accelerating within the installed base. So I can't provide a metric today regarding the percentages that use all three; there's a fair number of customers. I'd say some customers are compute-only, but broadly speaking, the majority are either a security customer or a compute customer. Roughly 74%, 75% of our customers are both security and delivery customers. So that's remained fairly steady across the years, perhaps risen slightly. But yes, we don't have those — necessarily those pool of funds deals or ELAs in which there's a large commitment without product specifically addressed or volume against a particular product line.
Gabriela Borges
analystI'd like to pick up right from where that earlier discussion ended. So regarding the possible stages of inference deals, Ed, I believe you casually brought up the rough scale of $50 million to $100 million. I'm curious to hear a bit more on that. Can the pipeline actually handle deals around that size, or is it already demonstrating them? Because I can picture how something like that might become a significant piece of your growth formula heading into next year.
Edward McGowan
executiveYeah. What I meant was, suppose you had a big — since we've encountered a few — parties interested in some rather big deals. If a deal that large came along, you'd end up purchasing additional CapEx beyond that amount. So suppose I received an order for 50 to 100, or roughly that, I'd likely be buying more CapEx, since greater demand would follow. Yet the real point I wanted to convey is that certain customers are approaching us about deal sizes that are quite substantial, bigger than what we normally encounter in compute these days. So I'd say, broadly speaking, the deals will definitely be bigger right from the start.
Gabriela Borges
analystYes, that's really fascinating. My next question, and this could go to either you or Tom, concerns the delivery segment. For a few quarters now, you've been reporting the stabilization and growth figures you've mentioned. Could you remind us whether there are unique opportunities tied to when cohorts reach their refresh point, depending on which year that falls in, that shape those pricing trends — and also remind us why pricing today is steadier than it used to be?
Edward McGowan
executiveSure, that's a fair question. Generally speaking, whenever we have — right now, we have 6 customers accounting for at least 1% of our revenues. So concentration risk isn't really significant for us. However, whenever those 6, or our top 10, hit their renewal dates within a single quarter or inside a 6-month window, that will certainly affect revenue. That situation didn't arise this year. We don't anticipate it happening next year either. What we have is simply the usual revenue renewal patterns we always see. We're encouraging and motivating our sales reps to secure longer contract terms, and we're seeing progress on that front, which is encouraging, including among some of our bigger clients, which is also encouraging. Regarding how pricing is behaving, we've observed some settling in the bigger media deals, where revenue tends to be impacted most. There's still a bit of pricing pressure across the broader base, so to speak, but it's nothing unusual — in fact, it's somewhat better than what we've experienced in the past. As for what's causing that, I'd say it's simply that there are fewer competitors around. Many of them have shut down or, in certain instances, simply left the space, even if they're an established company that isn't solely focused on delivery. So that could be playing a role too.
Frank Louthan
analystOverall, delivery has had a fairly strong year. What's really behind that, and is there anything early in the Q3 trends that hints at how seasonality is taking shape? Also, with the bigger contract deals you're signing with customers, is there any risk when it comes to sales cycle timing?
Edward McGowan
executiveYes. Typically — so with longer sales — sorry, I keep saying cycles — I mean longer contracts, no, it doesn't always stretch out the sales cycle when you're dealing with existing customers; for newer customers, clearly, your sales cycle runs a bit longer, but not for existing customers, no; as for how the business behaves, Q3 is generally somewhat softer as a seasonal quarter for delivery, simply because of content dynamics, sports seasons, vacations, things like that. So we observed what looked like a typical pattern there. But traffic growth is likely the main factor, paired secondarily with pricing declines having eased steadily across the year. So it's truly a mix of those two things.
Sanjit Singh
analystNice work on those strong numbers. I'd like to revisit the compute side of things, and it's encouraging to see CIS picking up speed. Regarding that additional $100 million within that area of the business, I realize there are challenges tied to the storage component, plus some video optimization closures, I believe, that you were moving over to your partner. When might that cease to be a drag on growth overall? I'd just like an update on that part of the reference.
F. Leighton
executiveRight. Those fall under our other cloud compute applications segment. That area isn't truly where the business is focused. A bit more will still occur there, but roughly speaking it came in flat versus the prior quarter; it isn't where our investment goes, and it isn't where growth will come from. My sense is you should zero in on CIS, Cloud Infrastructure Services — that's where the energy is, and that's where all the future growth opportunity lies. In fact, looking ahead, we might even split the two apart and report on CIS, since OCA is fairly small and fairly steady; the major force behind our future growth is Cloud Infrastructure Services, and that's where our offerings sit, such as Edge Workers for Function as a Service, along with 4,000 POPs, and the manage container service that a hyperscaler was using for their media workflow and AI. Then there's Akamai inference Cloud, which holds enormous growth potential. All of that belongs to Cloud Infrastructure Services. So that's likely what you should keep in mind when you consider Akamai Compute.
Sanjit Singh
analystRight. And that's the same thing we're examining too. I'd simply like to understand when — perhaps when a portion of those revenue headwinds begin to diminish. But that's reasonable. Regarding the Inference Cloud opportunity, what are you kind of picturing? Is there a chance for security to be attached to Akamai Inference Cloud, and I'd really like to hear — I'm curious how you're approaching the security side of Akamai Inference.
F. Leighton
executiveYes, definitely. And that is an excellent question, since when you're deploying these applications and models, they must be protected; actually, they carry even bigger vulnerabilities than a typical application or API would. So you're going to require API security, you're going to require something like an AI firewall; and those are areas where we possess market-leading solutions, so we can build that surrounding your inference engines and your models that are on Optimize Cloud.
Fatima Boolani
analystTom, there's a broad-picture item I'd like to raise with you. Let's consider how traffic is composed — say, the whole traffic footprint flowing across your pipe. How do you see the split among traditional video delivery and OTT on one side versus enterprise on the other, and perhaps now even AI plus certain consumer-facing AI uses that could add lift there? What I'd like to know is how you're viewing — well, where matters stand at present when it comes to the traffic mix, given AI and bots growing. And how ought that shift across the coming 2 years, and how does it connect with your delivery and CIS business? I'd appreciate your take on this. Then I have a follow-up for Ed, please.
F. Leighton
executiveYes. That's an interesting question. At present, most of the traffic consists of video — whether on-demand or live — along with software downloads; now when you look ahead, you're correct, sites that — commerce sites that previously might have carried no video or very little, they're already considering how to make the whole experience immersive through video. So when you land on a page, it opens with a video of you wearing something they believe you'd want to purchase. And that does push traffic higher. Beyond that, even conventional video, live sports. As we've discussed, we already have a customer where — the feed I get while watching the game will differ from the feed you get or such — it will be tailored to the individual; and so AI will be applied even for watching conventional video to deliver a better, more personalized experience. And who knows whether people ever end up wearing devices that provide augmented reality, well, that generates plenty of additional traffic; so I do believe you're correct that there's a reasonable prospect that as AI grows more widespread it will affect the traditional delivery business in a manner favorable to Akamai.
Fatima Boolani
analystAnd Ed, this one is a fundamentals-type question, building on part of what you already went over; could you take us through how the business looks at a high level for gross margin and gross profit when broken out by segment. And really, what I'm getting at is you've delivered really strong operating leverage in the business even though I would have assumed the delivery segment perhaps carries the weakest gross margin profile compared with the other segments, or revenue segments rather; so I'd love a quick recap of how those gross margin profiles compare and what some of the puts and takes are, particularly regarding some of the CIS business and the CapEx needs you mentioned earlier.
Edward McGowan
executiveYes, sure. So I'd say, a couple of years back we had shown some margin across the 3 segments including security; and looking at gross margin, roughly high 80s. Compute sat in the low 70s and delivery is probably high 60s, if my memory serves. It's likely still around that same area. And I think generally the business will operate at roughly low 70% gross margins, maybe slipping a point or 2 as we build out scale for — there's a bit of a timing mismatch between demand, revenue, and the build-out, since the build-out has 2 components. There's the colocation where you're — particularly with AI Inference you'll be making more — bigger colo purchases, so potentially you get some unfavorable accounting. So compute gross margin might drift down somewhat for a while. But we do believe you can reach roughly that low 70s gross margin. Looking at CapEx, even including AI Inference and what we sell with the AI Inference Cloud, that dollar of revenue to dollar of CapEx is a fair proxy for what we're observing; and as I mentioned, it might be somewhat better with GPU as a service since there's scarcity there. So there's a fairly solid market price established for that right now, though it could be somewhat better; so I do think that, on the whole, you get solid leverage because our sales force, the way we'll structure it, you have some specialists, but that's reasonable in terms of the specialists, and our sales force can sell everything and pull in specialists, so we ought to get solid operating leverage throughout the whole business. So it isn't only delivery that provides solid operating leverage, which it does, but all 3 products will provide solid operating leverage.
Jonathan Ho
analystWhen it comes to CIS, would you say these are primarily new deals for compute and capacity that you're landing? Or does some of the growth also come from workloads that already exist and are being moved over to CIS? And on a related note, could there be a chance to gain from AWS and Azure outages, since customers are seeking more resiliency?
F. Leighton
executiveIt's both. But the point you bring up about outages matters a great deal; and it's something people don't give much thought to. And this holds true whether someone is purchasing compute, delivery, or security as well. Actually, it might matter even more when it comes to security. At Akamai, what we aim for is five 9s of reliability. What that means is across 2 years, the site or app experiences downtime of under 10 minutes in total from every cause, attacks included; and we hit that mark right now, and we're certain of it since some of the biggest banks on the planet, along with regulators, demand Five9; and we've poured an enormous amount of investment into our platform so that whenever we push out metadata or software, or a customer pushes out their metadata, there isn't some unintended result that takes down the platform. And we watch what happens to our rivals who skip that investment. Take, for instance, security firms we go up against where a number of them, every quarter they're offline for an hour, and for a bank that's simply catastrophic. In the world we live in now, they can't make it that way. So reliability really matters. And that's something Akamai is genuinely strong at. So I'd say that's incredibly important. Great point to bring up.
Jonathan Ho
analystAnd then perhaps to build on that. Regarding the security side of your company, could you share some detail on adoption levels for areas like segmentation, plus how things are going with the AI security offerings?
F. Leighton
executiveSure. So plenty of upside remains in segmentation and API security; and further, like we mentioned before regarding API security, an entirely fresh market is emerging around AI inference engines, models, and agents, all of which will require specialized protection. Clearly, it's still early for our particular AI firewall safeguard, yet the customer base has shown very keen interest, with hundreds of prospects and numerous proofs of concept, and the first customers are now beginning to adopt the solution; and my sense is that this landscape will evolve quite rapidly when it comes to safeguarding the models, the inference engines, and the agents.
Will Power
analystAll right. Excellent. So, I suppose, two items. Tom, perhaps we begin with you. It would probably be helpful — and I realize you've already addressed this to a decent extent. When I consider the Akamai Inference Cloud plus that NVIDIA alliance, could you talk about what it means strategically over the medium and longer haul? Is the main thing simply getting hold of black well GPUs? Or do other dimensions and elements of this collaboration matter — anything regarding go-to-market, and so forth? After that, I've got something to ask Ed.
F. Leighton
executiveWhat I'd say matters most is that this delivers quite robust compute power for AI in proximity to users, data, devices, robots, and agents — that's the distinctive part of what we're offering. Essentially, you're putting together Akamai's distributed platform — by a wide margin the most distributed anywhere globally — alongside top-tier GPU capacity. Those new 6,000 units pack serious power. Models containing hundreds of billions of variables can run on them. They enable genuinely impressive things that simply weren't achievable previously, such as displaying to a visitor who arrives at the site, within the site, in a video. Really neat capabilities. And I believe that's something only Akamai can do; also, our relationship with NVIDIA is solid, and the two of us are collaborating when it comes to commercializing this capability alongside customers and the ecosystem.
Will Power
analystAll right, fantastic. And Ed, I suppose impressive improvement on the operating margin front, and it seems like you largely anticipate that carrying on through Q4. So it would be helpful to hear anything about what's behind that improvement versus what had been anticipated. And regarding this 30% adjusted operating margin figure, is that roughly the correct way to view things as we head into the following year?
Edward McGowan
executiveYes; guidance for the following year will come in February. What I'd point to is execution, which I brought up at the start of the call. A portion of that is generating extra revenue. Another portion is simply — even considering our development work and capitalizing labor, which serves as a solid productivity measure; that came in somewhat higher during this quarter; thus we got more done. The team has performed well even on procurement with certain of our suppliers when it comes to securing improved pricing. We're upgrading portions of our back office, which lets us shut down some legacy systems and move everything onto a single one. Oracle is a solid illustration of what we're moving everything onto. We've had excellent execution from our team handling the network build-out when it comes to securing improved pricing for co-location and bandwidth and those kinds of things. So what I'd say is it's simply solid execution in every area.
Patrick Colville
analystI'll try to squeeze in a quick one for Dr. Tom, then one for Ed if there's time. So Dr. Tom, my question concerns the NVIDIA partnership — there were a few announcements in the prior year, and Akamai bringing NVIDIA GPUs to the Edge. Could you clarify what's different in late 2025 compared with the prior year under this NVIDIA partnership?
F. Leighton
executiveThe relationship goes far deeper now. We're currently rolling out the Blackwell 6000, which represents an enormous step up in capability from the 4000 we had before, and we're putting them into service across a far wider footprint, plus the partnership and relationship are considerably stronger; so it's an entirely different situation, and that's what makes the Inference Cloud possible along with all the applications we discussed.
Patrick Colville
analystAll right, that's very clear. And congrats on that partnership — exciting news. Ed, if I could, a quick question about the numbers. The first person on the line basically asked about the soft guidance you gave for each business line. If I heard you correctly, I believe you said 10% for security, and just under 15% for Compute. So if I've got that right, it suggests delivery probably steps back down again in 4Q. Is that the correct way to look at it? And if that's the case, does that reflect conservatism? Or is there a reason growth in delivery would turn back down that we ought to be aware of?
Edward McGowan
executiveRight, we gave a pretty broad range there, and Q4 delivery is genuinely hard to call. So please don't read that as us expecting something bad on the delivery side. You really don't see the quarter's seasonal shape kick in until roughly Thanksgiving, which makes it hard to say how the retail season plays out or how the year-end media season looks as new devices come online. That leaves you with a spread of possible results this quarter. Right, so my comment was that Security ought to land around 10%, Compute slightly under 15%, though that's actually a promising sign looking ahead to next year, since the whole thing comes down to timing, with delivery essentially making up the difference. So your results for delivery will vary quite a bit based on where in that range you set your model.
Andres Miranda Lopez
analystAndres Miranda here, filling in for Rudy. Going back to how the call started, you mentioned that CIS is now in use at the 3 bigger cloud providers. To us, that suggests it wasn't true previously. Could you simply verify that? And assuming it is true, did you sign just one of them during this quarter? Also, what opportunity lies ahead? In what way should we view the 3 large cloud providers.
F. Leighton
executiveF. Leighton, A: I didn't fully catch what you asked. Would you mind saying your question again?
Andres Miranda Lopez
analystYeah. So early on in this call, you mentioned that three of the bigger cloud providers are now using CIS. Does that mean that wasn't true previously? So did you only bring one of them on board during the quarter? Or what's the opportunity ahead for CIS?
F. Leighton
executiveRight. We brought the third one on — the third one was just signed. And the first one signed a contract that's a lot larger. So a single one grew substantially bigger, and a single one is new. And so at this point all three are with us. Is that what you meant?
Andres Miranda Lopez
analystYes. Thank you.
Tomer Zilberman
analystI'd like to circle back to Security and discuss what you're observing on the demand side. When I listen to what some of your competitors are saying, they describe an underlying demand environment that's picking up speed. Yet looking at your own numbers, growth has been running around 9% to 10% over the past several quarters, and you guided to 10% for the full year. So I was hoping to get the pluses and minuses behind what you're seeing on demand, and how you're framing that as you head into next year.
Edward McGowan
executiveYes; so I believe the key thing we want to emphasize today is that demand is substantial for our API Security offering and our Guardicore Zero Trust platform security products. With API Security specifically, going back a bit, we completed the Noname acquisition in June. That makes this the first quarter where we have full organic year-over-year growth. And we more than doubled. So we'll exit at roughly a $100 million run rate. So demand for API security is incredible, and Guardicore continues to see strong demand; so — and there's also — as Tom mentioned, with API Security and what's taking place with the Agentic Web and the chance for growth ahead, looking forward, API Security has considerable runway, and the penetration rate within our customer base for both of those products remains fairly low, and we've demonstrated a strong record of achieving very high penetration. So a few of our products have existed for quite some time and are still expanding, though not at the same pace; so I think you'd be overlooking the opportunity if you only looked at the total figure. You really need to examine the two rapidly growing products, the demand underlying them, and our standing in the market — we hold a very strong position there with market-leading offerings.
Daniel Hibshman
analystDaniel Hibshman here, filling in for Jeff Van Rhee. Nice work on the quarter. Perhaps to start, on the delivery trajectory and what's anticipated there. Is this a line where, as ever, the growth rate is constantly shifting purely depending on that year's market dynamics? Or do you view the trend line we're settling around here — roughly single-digit growth, or a single-digit decline — as the trend line to expect going forward? Just your long-term vision for how that behaves secularly.
Edward McGowan
executiveYes. Fair question; there's seasonality in Internet traffic trends no matter what. The fourth quarter is invariably a bigger seasonal period. College football and the NFL come back, plus there's fresh content for new TV shows and that kind of thing. So that invariably pushes more traffic as new devices come online. You also have a huge holiday shopping season. So a traffic jump there is very typical. Usually, the first quarter might come in somewhat below what you saw in the fourth quarter. The second quarter generally aligns with the first quarter. Then the third quarter tends to be seasonally a bit softer, just traditionally, since not as much sports is happening. You don't have this — no fresh shows are launching and people take vacations, so they aren't online as much. I don't expect that broader online pattern to change direction. There's a chance AI agent apps wind up driving additional video and API traffic, as Tom pointed out, which could give things a boost. Over the past two years, gaming has been soft, and a number of major titles remain unreleased. That's a positive factor, and so is the console upgrade cycle — one of those should land within the coming one to two years. So my view is that demand could strengthen as time goes on. We've still been fairly careful when it comes to delivery, since surprises can pop up now and then. That said, what we're seeing makes us quite happy, and we're sort of returning to the point where our projection for the business had been roughly flat to a low-single-digit decline — and that is definitely the result we turned in for this year. Should conditions hold steady the way they did over this year, there's a chance we might repeat that outcome.
Daniel Hibshman
analystThen there's one more thing I wanted to ask about regarding share repurchases — is there any change in methodology or strategy worth mentioning, both when it comes to those record buybacks we saw during the first half, and then, well, I was just curious and took a look, and it seems like this is the first quarter. It appears Akamai hasn't repurchased any shares since roughly a quarter back in 2009. So normally there's just a really, really consistent cadence. Is this somewhat of a move toward purchasing more opportunistically in more concentrated amounts? Or any thoughts at all on whether there's any change in the thinking there?
F. Leighton
executiveNo. Like Ed mentioned earlier, there's no change in how we think or our strategy. And actually, we've bought back more shares, or spent more money on shares, this year than at any point before; so yes, during Q3 we didn't buy back shares, but this year we bought back a ton.
Jackson Ader
analystOver the past two quarters or so, the gap between how we talk about international and the U.S. has narrowed somewhat. I'm just wondering: within the international segment, which areas came in strong, or perhaps even better than you'd anticipated? And thinking historically, when international outperforms the U.S., what segments ought we to keep top of mind?
Edward McGowan
executiveYeah, fair question. You've got it right — it did narrow somewhat. One factor that helped U.S. growth during last year, or '25, was that a number of the Edgio contracts skewed more heavily toward the U.S.; now those are beginning to anniversary, so that plays a role. On the international side, though, I'd point to APJ specifically, where the sales team has really taken to compute, in a sense driving the effort when it comes to inking many big compute deals, and taking on security too. So there's strength pretty much everywhere there. APJ overall has performed extremely well for us. EMEA has done well. A few of the country-level growth figures I've observed throughout Western Europe were somewhat unexpected — healthier, actually, than what we typically see. So I'd simply say that internationally, things have always gone well for us. We're watching it regain momentum once more. And I'd say compute plus APJ is likely the item I'd single out when it comes to strength.
Jackson Ader
analystAll right, that's helpful. Now, a fast second question about Inference Cloud. Tom, you brought up how you sit near the robots—near the user—and I get that. But what I want to know is whether other developments beyond what I'd basically call Internet of Things scenarios could drive additional inference at the edge. For instance, if small language models took off, or if silicon became more varied. Are there any headlines that should act as word associations for us—ones that would signal something good for Inference Cloud whenever we come across them?
F. Leighton
executiveYeah. I'm not sure I can give you buzzwords to watch for, but essentially everything will run on AI. The way the web functions is set to shift. Everything becomes model-driven, and the manner in which you engage with it won't be the same. Applications will emerge that nobody has imagined so far. Many of these—particularly where a bot handles something critical, such as a car, or where humans engage in some fashion—will require the response to happen live; what I mean by that is inside 10, a small count of tens of milliseconds. To achieve that, proximity matters, and you won't want to travel a long distance to a massive data center housing an enormous model that handles a zillion tasks, with plenty of capacity and infrastructure that has no real bearing on your test. So my belief is that you'll witness small- to medium-sized models spreading widely, running near the user while focused on particular tasks; that, I think, explains why industry leaders have begun describing Inference as—well, truly the coming major wave, and in the end bigger than the huge models, the training, and the Internet's core; and that this is where the Internet and the web are headed in the future.
Operator
operatorAppreciate it. That brings us to the end of the Q&A portion and also wraps up the call for today. Thanks to everyone for joining us for today's presentation. At this point you're free to hang up, and we hope the rest of your day goes great.
Read the full transcript via the API
You're viewing the first half of this call. Get the complete Akamai Technologies, Inc. transcript — plus 133,000+ transcripts from 7,000+ companies, speaker segments, AI summaries and full-text search — through the EarningsCalls.dev API.
Get the API View API docs →This call discussed
For developers and AI pipelines
Programmatic access to Akamai Technologies, Inc. earnings transcripts and 133,000+ others is available through the
EarningsCalls.dev REST API. Plans from $39.99/month — full transcripts, speaker segments,
full-text search, and the recently-added /api/v1/transcripts/recent polling endpoint for ETL pipelines.
More tools for stock research
OptionsBell: an unusual options activity scanner that tracks options flow across all US stocks and emails you when large trades appear in Akamai Technologies, Inc. or any ticker on your watchlist.
StockMarketScan: a stock screener with alerts: 24 pre-built screens refreshed every 5 minutes, watchlists and email notifications. Both offer API and MCP access and fair monthly pricing.