Akamai Technologies, Inc. (AKAM) Earnings Call Transcript & Summary
November 5, 2026
What were the key takeaways from Akamai Technologies, Inc.'s November 5, 2026 earnings call?
In Q3 2026, Akamai Technologies reported revenue of $1.122 billion, a 6% increase year-over-year, with a non-GAAP EPS of $1.77, down 5% year-over-year. The company announced a significant $11.6 billion commitment with Anthropic for cloud infrastructure services (CIS), contributing to a total of over $14 billion in multiyear contracts signed this year. However, management indicated that supply chain delays would result in a revenue shift of approximately $30 million from Q4 2026 to 2027, leading to a lowered revenue outlook for Q4. Despite this, Akamai raised its 2027 revenue guidance to $1.1 billion to $1.3 billion for CIS, projecting overall revenue growth of 17% to 22% in constant currency over 2026 levels.
What topics did Akamai Technologies, Inc. cover?
- Cloud Infrastructure Services Growth: Akamai signed an $11.6 billion commitment with Anthropic, bringing total CIS contracts this year to over $14 billion. Management stated, "We now anticipate that our CIS revenue will more than double in 2027 to $1.1 billion to $1.3 billion."
- Revenue Shift Due to Supply Chain Issues: Management acknowledged that supply chain delays would shift approximately $30 million of revenue from Q4 2026 into 2027, stating, "These issues have since been resolved, and they won't impact our outlook for 2027."
- Increased Profitability Expectations: Akamai expects the large CIS deals to contribute to an aggregate adjusted EBITDA margin of about 75% and a non-GAAP operating margin of approximately 30% when fully ramped. Management noted, "We believe that the large deals we've signed this year will contribute at an aggregate adjusted EBITDA margin of about 75%."
- Security Revenue Growth: Security revenue grew 9% year-over-year, driven by demand for web app firewall and API security solutions. Management highlighted, "AI continues to be a tailwind for our security solutions as more AI agents act on behalf of users."
- Future Revenue Guidance: For Q4 2026, Akamai projects revenue between $1.145 billion and $1.170 billion, representing a 5% to 7% increase year-over-year. The full year 2026 revenue guidance is now set at $4.4 billion to $4.465 billion, up 6% year-over-year.
What were Akamai Technologies, Inc.'s November 5, 2026 results?
- Revenue: $1.122 billion (+6% YoY)
- Non-GAAP EPS: $1.77 (-5% YoY)
- CIS Revenue: $109 million (+34% YoY)
- Security Revenue: $622 million (+9% YoY)
- Q4 Revenue Guidance: $1.145 billion to $1.170 billion (+5% to +7% YoY)
- CIS Revenue 2027 Guidance: $1.1 billion to $1.3 billion
Akamai's strong pipeline and significant contracts in the CIS space position it well for future growth, despite near-term revenue shifts due to supply chain delays. The raised guidance for 2027 reflects management's confidence in the long-term potential of its cloud infrastructure services. Investors should monitor the execution of these large contracts and the impact of macroeconomic factors on growth.
Earnings Call Speaker Segments
Mark Stoutenberg
executiveGood afternoon, everyone, and thanks for being with us on Akamai's earnings call for the third quarter of 2026. On the line today are Tom Leighton, who serves as Akamai's Chief Executive Officer, and Ed McGowan, Akamai's Chief Financial Officer. I want to point out that our remarks this afternoon contain forward-looking statements, including guidance on revenue and earnings. Such statements reflect our present expectations and assumptions, which are subject to specific risks and uncertainties and depend on various factors that might lead actual results to differ materially from what is expressed or implied. Those factors include, though they aren't limited to, effects arising from macroeconomic trends, the integration of any acquisition, geopolitical developments, and the additional risk factors described in the filings we make with the SEC. What we say on this call reflects the company's views as of November 5, 2026, and we undertake no obligation to update any forward-looking statements. As a reminder, during today's call we will be referring to certain non-GAAP financial metrics. You can find a detailed GAAP to non-GAAP reconciliation in the investor relations section of akamai.com under financials; with that, I'll now turn the call over to our CEO, Dr. Tom Leighton.
Tom Leighton
executiveThanks, Mark. Akamai maintained strong momentum in expanding Cloud Infrastructure Services, or CIS, throughout Q3. The standout development: we secured an $11.6 billion commitment running seven years with Anthropic, through which our distributed infrastructure and software will handle their CPU workloads. That brings cumulative multi-year CIS commitments signed this year to more than $14 billion. That total includes an earlier substantial Anthropic contract, two major agreements with leading AI companies, and numerous multi-million dollar deals with large enterprises around the globe. I'm pleased to report we're performing well against these contracts. Revenue is already flowing from two of the first three large deals, and we anticipate the third will begin generating revenue later this quarter. With two of those deals, we encountered modest supply chain and third-party data center delays, meaning our Q4 revenue outlook will land below prior expectations; those issues have been resolved, and they won't impact our 2027 outlook. In fact, we now project CIS revenue will more than double in 2027, hitting $1.1 billion-$1.3 billion, and consequently Akamai's overall revenue growth rate should speed up dramatically, with total 2027 revenue rising 17%-22% in constant currency compared with 2026 levels; we anticipate further acceleration in 2028, when overall revenue should surpass $7 billion. As CIS becomes a larger portion of our revenue, we anticipate a strongly favorable impact on margins and EPS growth. To be specific, we believe this year's large deals will, once fully ramped, produce an aggregate adjusted EBITDA margin of roughly 75% and a non-GAAP operating margin of approximately 30%. Several factors underpin our confidence that we can operate CIS at strong margins at scale. First, it's essential to understand that bare metal isn't what we're discussing here. In fact, we don't offer bare metal as a service at all. What we sell is compute capacity through our infrastructure as a service platform — virtual machines, managed Kubernetes, storage, and networking; that keeps capacity elastic across customers and generates substantially more revenue per megawatt or per dollar of CapEx than a bare metal contract would; specifically, for this year's large deals, we expect roughly $22 million in revenue per megawatt of contracted power. Moreover, we're seeing that as our market presence expands, the pricing we can command has strengthened. Indeed, unit pricing in recent deals is far higher than in comparable deals signed earlier this year; we also benefit from several cost advantages owing to our expertise and the distinctive nature of our cloud platform. For example, we operate one of the world's largest backbones — larger even than many tier one carriers — which reduces our networking costs and enables us to leverage scarce data center space that others may be unable to access, with revenue projected between $3.462 billion to $3.62 billion. Since our platform spans the world, we can place compute across many smaller locations instead of concentrating it in just a few massive data centers. That's another factor enabling us to obtain the power and colo capacity required for expansion, and to get it at lower cost than we would by depending on very large footprints in expensive core data centers. For a long time, we've been honing our software and operations so that we can process large volumes of customer data with greater efficiency. In terms of moving data over the internet, we rank among the largest players, since most of the world's biggest brands — including several hyperscalers — depend on Akamai to secure and deliver their content. A still more important factor in our ability to sustain healthy margins at scale is the type of servers we operate in our cloud. We do offer a range of GPUs to customers, but the vast majority of the servers we deploy — well over 90% — are CPUs; in fact, mostly the same CPU type we use for our security and delivery services. That's significant because CPUs are being used much more widely for AI. The reason is that, compared with GPUs, they tend to be more versatile, more cost-effective, and longer-lasting. The CPUs we deployed in the past continue to deliver value across our platform, even seven to 10 years afterward. In fact, the largest cloud contracts we signed this year guarantee revenue for those CPUs across seven years. Using the same servers for compute, security, and delivery stretches their useful lifespan even further and makes our platform more efficient at the same time. CPUs typically consume far less power than GPUs, so the megawatts needed to power our cloud are far fewer than for providers such as the neoclouds, who rely more heavily on GPUs. That means we spend more CapEx per megawatt than the neoclouds do, but for Akamai shareholders that's a good thing. Power is scarce and costly, and relative to our CapEx spend we need less of it than others need for their infrastructure. There are plenty of reasons we believe our fast-growing CIS business will be very profitable for Akamai, but probably the biggest one is that we offer a compelling and highly competitive solution in the cloud marketplace. This doesn't just support profitability — it also helps ensure Akamai will keep growing durably, even if resources that are scarce today, such as power, equipment, and data center space, become more abundant down the road. There are many reasons the world's most important companies pick Akamai for their AI and cloud computing needs. One of the biggest is that, with more than 4,000 points of presence across 130 countries, we can bring customers' compute workloads closer to users and devices wherever they are on the planet. Having that proximity lets us deliver reduced latency and greater throughput, and that matters across many use cases—interactive video, avatar rendering, conversational voice agents for customer service, speech recognition, translation, robotics, video processing, and real-time alerting. That is a big reason why even the hyperscalers turn to Akamai Cloud for some of their latency-sensitive applications. Another substantial advantage for Akamai is that we can bundle compute offerings together with our market-leading security and delivery solutions. On one platform, enterprises are able to power, protect, and deliver their most critical applications with excellent performance and tremendous global scale. Cross-sell benefits work both ways as well. We can offer compute to our sizable base of security and delivery customers, and on the flip side, we can offer our security and delivery solutions to new cloud customers. Protecting customer workloads matters especially in the AI era. Not only can we help shield customers from AI-enabled attacks, but we can also help keep their agents from going rogue and causing unintended harm. Reliability is another key advantage for Akamai when enterprises choose a cloud provider. As a strong proof point, we let the world's largest financial institutions hit five nines of uptime—equivalent to under 10 minutes of downtime over a two-year span, with revenue ranging from $3.462 billion to $3.62 billion. Reaching that is no easy task.
Tom Leighton
executiveHonestly, pulling that off is really difficult — it demands exceptional skill, commitment, and hands-on experience. On top of that, Akamai delivers best-in-class professional services and customer support, backed by thousands of highly knowledgeable technical specialists who assist our clients with their toughest problems. The assistance we give customers is a major factor in why we take business away from competitors. Over 25 years, Akamai has earned a solid reputation as an innovative, dependable partner. We are no newcomer. We bring decades of experience guiding the biggest and most exacting enterprises through a technology landscape that keeps shifting fast. On the innovation front, only last week we broadened our cloud portfolio by launching Akamai Inference and Akamai Agent Sandbox; these services let enterprises build and tune open-weight models using their proprietary data and run agents inside strictly enforced boundaries, all secured and powered on Akamai's AI-optimized cloud platform. They supply the inference and execution layers that transform our AI infrastructure into a full AI stack supporting training, real-time inference, and agentic workloads around the globe. Akamai Inference and Agent Sandbox also connect with our Firewall for AI and new agent security services, ensuring our customers' AI remains safe. Now on to security: I'm happy to share that security revenue for Q3 rose 9% year-over-year as reported and 10% in constant currency. Once again, growth was driven by robust demand for our market-leading web app firewall, API Security, and Guardicore Segmentation solutions. We were particularly glad that industry analysts at IDC and Forrester named Akamai a leader in all three of these product areas in reports published last quarter. AI keeps acting as a tailwind for our security offerings as more AI agents act on behalf of users and enterprises, as more employees place sensitive data into AI tools, as more attacks leverage AI to expand in scale and sophistication, and as more AI agents and bots become better at seeming human. Akamai security solutions help companies shield their applications, infrastructure, and workforce from AI-related vulnerabilities and attacks. For instance, our web app firewall blocks malicious traffic before it reaches an application, and it patches exposure points automatically. Our Bot & Agent Control solution unmasks agents to reveal their true origin and intent, then takes the right action on behalf of our customers. Our API Security solution detects and defends shadow AI and APIs before sensitive data can be stolen. Our Guardicore Segmentation solution limits the damage if and when something bad does get through. To sum up, we believe Akamai offers something unique in the cloud marketplace, with superior performance and security, global scale, and strong financials. Our distinctive cloud platform lets enterprises create and expand the next wave of AI applications — everything from model training at the core to rapid inference at the edge — all safeguarded by Akamai's industry-leading security solutions and scaled through Akamai's industry-leading delivery services. Because of this, we think we're ideally situated to gain as businesses embrace AI at scale. Already in this year, we've signed deals to power and protect mission-critical applications for the world's top frontier model companies, top AI equipment companies, major open-source model companies that are building the next generation of AI applications, and the world's top enterprises. No matter who ends up being the dominant AI players and however the AI landscape shifts, we believe that the breadth of our AI customer base, together with our very strong enterprise relationships and our distinctive cloud platform, will generate substantial growth for Akamai well into the future. My confidence in Akamai's ability to power and protect an AI-driven world has never been greater. Now I'll hand the call over to Ed, who will walk through our financial model and outlook. Ed?
Ed McGowan
executiveThanks, Tom. Before I review the Q3 results and walk through Q4 guidance, I'd like to build on a few points Tom made about Cloud Infrastructure Services — in particular, the $14 billion in multi-year CIS commitments that were signed and announced during this year. First, let me address some questions investors have raised about the profitability and return on investment tied to these customer wins. Today we posted a slide in the quarterly results section of our investor relations website that shows the aggregate returns we expect once the full $2.2 billion of annual revenue from these $14 billion of signed contracts has completely ramped and is being recorded in our results; the slide also presents our calculations for an expected after-tax internal rate of return of approximately 13% and an expected non-GAAP average annual adjusted return on invested capital of approximately 15% on these deals. As the slide shows, once fully ramped we expect aggregate total revenue of approximately $2.2 billion a year, along with an aggregate adjusted EBITDA margin of approximately 75% and a non-GAAP operating margin of approximately 30%. As we've said before, we expect this revenue to start ramping in Q4 2026 and to be fully ramped by the end of 2028. Second, as Tom noted, we have started recognizing revenue from two of the first three major CIS contracts, which represent $2.8 billion in total contract value. The third contract is in its final testing phase, and we expect it to begin generating revenue soon. As Tom mentioned, revenue on two of these deals has been delayed because of some supply chain issues and third-party data center delays. I'm glad to say those issues are now resolved. Because of these delays, we expect roughly $30 million of revenue to shift from Q4 into 2027. To be very clear, none of these timing delays affect the total revenue over the contract period, nor do they result in any penalties. They only adjust when revenue is recognized across periods. Next, on the recent $11.6 billion deal announced a few weeks ago, we're pleased to confirm that we have secured and contracted for all the space and power — colo — needed to deliver the full contract value, and we expect these sites to be available to us throughout the revenue ramp period we outlined earlier. We've also secured approximately 70% of all the hardware, memory, and other components required to deliver on this contract, including $1.7 billion of memory and other critical items in Q4 2026. As a reminder, this $11.6 billion contract includes a pricing adjustment mechanism designed to address any possible price changes in the hardware, memory, and other components that have not been ordered. So our margin is protected if prices rise between now and when we place orders for the remaining hardware and other components. So to wrap up and look forward, for 2027 we're expecting CIS revenue between $1.1 billion and $1.3 billion, and that should help push total company revenue growth into the 17% to 22% range in constant currency year-over-year. Then the year after, we expect total company revenue growth to speed up even more, exceeding $7 billion in 2028. It's worth noting that this $7 billion expectation for total company revenue in 2028 assumes no additional orders from Anthropic beyond the $11.6 billion we announced in late September — in other words, we're not assuming any revenue from the $9 billion option tied to the Anthropic relationship. So in closing, we have a lot of confidence in our ability to deliver the revenue we laid out today. Quarterly revenue might move around a bit versus the initial schedule on some of these large contracts, but any timing shifts won't affect the total revenue or profitability across the life of these agreements. Now let's move on to Q3. Third quarter total revenue came in at $1.122 billion, up 6% year-over-year as reported and up 7% in constant currency. Cloud Infrastructure Services, or CIS, revenue was $109 million, which is 34% year-over-year growth as reported and 35% in constant currency. Security revenue kept up strong momentum, coming in at $622 million, up 9% year-over-year as reported and 10% in constant currency. Our fast-growing API Security and Guardicore Segmentation solutions continued to drive the strength in the third quarter. Turning to delivery and other cloud applications, revenue was $391 million, down 3% year-over-year as reported and in constant currency. International revenue reached $563 million, up 7% year-over-year, or up 9% in constant currency, which is roughly 50% of total revenue in Q3. U.S. foreign exchange fluctuations hurt revenue by $1 million on a sequential basis and — $7 million on a year-over-year basis. Now on to profitability. In Q3, non-GAAP net income was $262 million, or $1.77 of earnings per diluted share, down 5% year-over-year as reported and down 2% in constant currency. Non-GAAP operating margin for Q3 was 25%, which is in line with our expectations. Q3 CapEx was $464 million, or 41% of revenue. Third quarter CapEx came in a little below guidance, mainly because equipment deliveries shifted into the fourth quarter. Turning to cash and our capital allocation strategy. In Q3, we didn't repurchase any stock. Year to date, we've bought back roughly 5 million shares for a total of approximately $616 million; we finished the third quarter with approximately $565 million left on our current repurchase authorization. As of September 30th, we held approximately $4.2 billion of cash equivalents and marketable securities. I'd like to touch on our current capital plan and funding outlook. Under our current plan, we expect to fund our CIS expansion and broader growth investments through 2028, as well as address our 2027 convertible note maturity, primarily through our existing cash balance and expected operating cash flows, which may be supplemented by measured incremental financing activities as appropriate. We stay very disciplined when it comes to allocating capital. If we do bring in outside capital, we'll look at structures that are flexible and cost-effective, with a sharp eye on protecting shareholder value. As our CIS customer commitments ramp, we anticipate a substantial rise in operating cash flow starting in 2028, and we believe that will strengthen our capacity to fund operations and investments through cash we generate internally while bringing leverage down over time. Ahead of giving our fourth quarter and full year 2026 guidance, I want to cover a few housekeeping items. First, a reminder that our Q4 2025 results had an unusually large $18 million contribution from security license revenue. We expect this year to come back to a normal level of $4 million-$5 million of license revenue in the fourth quarter. Second, on Q4 operating expenses, as is typical for our business, we expect them to be higher than in Q3. The main driver is the seasonal jump in sales commissions, as most of our successful reps achieve their annual quota accelerators; in addition, in Q4 this year, we expect additional margin pressure this quarter, driven by the ramping of CIS expense, mainly co-location costs, ahead of revenue.
Ed McGowan
executiveThird, we're still on schedule to invest roughly $1.7 billion during Q4 2026 in order to secure and pre-purchase key supply chain components — memory in particular — supporting the large Anthropic contract we told you about a few weeks ago; that spending will push our CapEx up meaningfully in Q4 and reduce interest income for the quarter. As a reminder, depreciation on that equipment starts once we begin recognizing revenue across 2027 and 2028. Last, looking ahead, we don't intend to announce each sizable CIS transaction unless SEC disclosure rules require it; notable commercial wins will still get highlighted in our usual quarterly narrative, and we'll keep investors informed as our capital deployment priorities and revenue dynamics keep changing. Now on to guidance. For Q4, our projected revenue range is $1.145 billion-$1.170 billion, which is 5%-7% higher as reported and in constant currency versus Q4 2025. At current spot rates, FX movements should affect Q4 revenue by — $4 million relative to Q3 levels, and by — $5 million year-over-year. Given those revenue levels, cash gross margin should come in at about 70%. Q4 non-GAAP operating expenses are projected at $368 million-$378 million. We anticipate a Q4 adjusted EBITDA margin of roughly 38%. Non-GAAP depreciation expense should be $165 million-$167 million; non-GAAP operating margin should be approximately 24%. With the revenue and spend picture I just described, Q4 non-GAAP EPS should land in the $1.50-$1.70 range; that EPS guidance assumes taxes of $48 million-$54 million, reflecting an estimated quarterly non-GAAP tax rate of about 18%; it also assumes a fully diluted share count of roughly 149 million shares. On to CapEx. For the reasons I mentioned earlier, we expect fourth-quarter spending of approximately $2.445 billion-$2.645 billion. Looking to full year 2026, revenue should be $4.44 billion-$4.465 billion, up 6% as reported and up 5%-6% in constant currency. Within Cloud Infrastructure Services, we now expect year-over-year revenue growth of 40%-42% in constant currency. Security revenue growth should still be in the high single digits on a constant currency basis in 2026. For delivery and other cloud apps, we still expect a year-over-year decline in the mid-single digits on a constant currency basis. At current spot rates, our guidance assumes FX will positively affect 2026 revenue by $5 million on a year-over-year basis. On to operating margins. For 2026, we estimate a non-GAAP operating margin of about 25%, measured at today's FX rates. Turning to CapEx. At this point, we anticipate full-year capital expenditures in the range of $3.462 billion-$3.662 billion. On to EPS. For full year 2026, non-GAAP earnings per diluted share should be in the range of $6.47-$6.68. This non-GAAP earnings guidance assumes an 18.5% non-GAAP effective tax rate and 148 million fully diluted shares. That covers everything on my side — Tom and I are glad to take your questions now. Operator?
John DiFucci
analystThanks. Ed and Tom, I appreciate all the details; this is excellent and genuinely useful. Last quarter, I believe you were clear that we shouldn't count on CIS speeding up in 3Q, and it was gracious of you not to rub it in our faces, yet somehow the consensus came in higher and indicated acceleration; this quarter, though, you largely delivered exactly what you'd promised. However, you cut your 2026 revenue figure, which you attributed to supply chain problems. Could you elaborate a bit more on the bottlenecks in the near term? On that note, we do value the long-term view, but how much conviction do you have in that 2027 CIS revenue figure, and could you shed some light on how it builds quarter by quarter through the year, if we take you at your word? Thanks.
Tom Leighton
executiveYeah, fair question; these were small sorts of issues involving the data centers. Think of it as a punch list on a construction project. It's not unusual, and we typically build in some slippage. This time they were worse than we had planned for. It's not a big deal. A couple months of delay. No revenue is lost, it simply shifts into next year. Our projections for the coming year are something we feel strongly about, and the ranges reflect that we've built in fairly large slippage should any occur. Should that not materialize, there's additional potential. Still, we have strong conviction about next year. On the ramp, the way to view it is that it builds significantly as the year progresses. Ed, would you like to add some detail on that?
Ed McGowan
executiveYeah. Hey, John. Thanks for the questions. Yeah, the way to think about it, I'll split it into halves for you, since quarters can shift a bit from one to the next. But figure on 25%-30% in the first half and 70%-75% in the back half when you're thinking about the ramping. Take the total amount and spread it out roughly that way. But I think it's important — you asked how confident we are, and we made some commentary on that. First, we're already beginning to recognize revenue. That's a big step. Second, we have all the colo we need for the large transaction we announced a couple of weeks ago. That's pretty significant, because as you know, colo is a tight market, but we have it all signed up and ready to go. We're just waiting for it to be delivered to us over the ramp period.
John DiFucci
analystThat sounds excellent.
Tom Leighton
executiveBeyond that, every approval we require is already in hand, which means the political worries some folks have won't ensnare us. For us, that simply isn't a concern.
John DiFucci
analystTom, I was actually about to ask that as my next question. Clearly there's a great deal of fear in the market at the moment, particularly regarding the build-out of AI infrastructure. So thank you for that ROIC slide — it's genuinely useful. But setting the math aside, just to be clear, you're essentially doing inference only at colo facilities. By only, I mean you aren't constructing those enormous data centers out in the world. This can be accomplished much faster than the big data centers that are used for training. So I suppose my thinking is — and tell me whether I've got it right or wrong — it appears the risk would be smaller. Execution risk remains, but the risk you can't control is smaller. Tom, you mentioned permitting. Could you expand a bit on any risk that's outside your control, such as power, real estate, and permitting, and how that affects your business compared with others who are mainly doing training and building out large data centers?
Tom Leighton
executiveThat's an excellent point, and these are very, very different businesses with very, very different risk profiles. We have already secured the power and space. The permits are already in place. Beyond that, we aren't discussing just one or two locations, where an unexpected event would deal you a big hit. With Akamai's approach and architecture, we operate in a lot of much smaller spaces. Should some location experience a small delay, it won't affect us in the big picture. It also makes obtaining the space much easier for us, since it comes in small chunks. It isn't tangled up in some permit for a nuclear power plant next door. You're right that we aren't in the business of training the big frontier models; we don't purchase that kind of equipment. It's an entirely different approach. As I mentioned, we're mostly doing CPUs; we do training for enterprise-sized models as well as post-training and inference. Most of the time, you aren't even using GPUs for that. This is where I believe much of the future market is headed, which is why we're concentrating on it. So it's far more profitable for us, with a much lower risk profile, and I truly don't see much execution risk in what we're projecting for next year. This is stuff we know how to do. We had a couple of months, okay, of a push. Not really a big deal, and I don't believe that has any impact on next year.
Samik Chatterjee
analystHello, and thank you for the opportunity to ask. I understand you've stated there won't be announcements about possible incremental deals or wins for the CIS business going forward, but could you perhaps share thoughts on the demand pipeline you're currently seeing? Tying that to what you mentioned regarding a capital raise, is it right to think a raise would happen only when tied to additional incremental wins? Or if you could shed light on how pipeline conversion should be viewed alongside a capital raise, that would be useful, and I have one more question after that.
Tom Leighton
executiveYes, our pipeline is quite robust, spanning hundreds of customers and encompassing some fairly sizable accounts, and we feel very upbeat about our ability to get that business closed. We've secured power and data center capacity to support it. As Ed mentioned, I believe we're well positioned given our capital and our capacity to bring in a modest amount of capital should we determine that's needed, doing so in a manner, consistent with how we've always operated, that pays close attention to shareholder interests.
Samik Chatterjee
analystUnderstood. I believe in your prepared remarks you noted that recent deals are coming in at a significantly higher premium than the $22 million per megawatt seen on some of the earlier recent deals. Could you help us gauge the size of that premium, and how that shifts the returns you're presenting in today's slides? What magnitude or gap are you observing in the market? And how does that affect the spread between the IRR and the cost of capital investors should anticipate? Thank you.
Tom Leighton
executiveYeah, I'll kick things off, and then Ed can walk through the specifics. But yes, on an apples-to-apples basis, pricing has improved. Part of that, I think, is that we're becoming much better known in the marketplace, which really helps. Ed, would you like to cover the details on the IRR and the economics?
Ed McGowan
executiveSure. The reason we shared those figures is that I'm aware there's been considerable discussion about how one would go about calculating this. So we supplied additional detail to let you run the calculation, and obviously those represent solid returns for us. Now, please bear in mind that these apply to the biggest volume and the lengthiest term. What's standard in the industry right now is that if you went out and bought something uncommitted, you'd be paying the top rate. So with no commitment and low volume, you get the highest rate. That means these represent the cheapest prices you'll find on a unit economic basis. So treat that as roughly your floor when it comes to pricing. If someone commits to half the amount or a third of the amount, a tenth of the volume, they won't come anywhere close to those prices. It's encouraging to see that, for the biggest possible deal, securing that kind of return across seven years is a very, very good business for us and for our shareholders. The deals now coming in are arriving at higher levels. Keep in mind that in this business, much of what drives unit pricing on the other side, on the buy side, depends on component costs. As those component costs rise, the market accepts higher prices. Our margins are protected by the fact that if co-location costs rise or if compute component costs rise, we're able to lift our pricing on subsequent deals; we lock in, as we discussed, these long-term deals; we have mechanisms in place to handle any pricing. Typically what happens is we've got the colo locked up for a long stretch of time, and we've got the equipment purchase; we're doing a really good job of protecting margins, and what you see there on those big deals is the lowest you'll typically see.
Jackson Ader
analystGreat, and thanks for fielding our questions, guys. Ed, if I could just piggyback on the margin topic. I understand the first half, second half revenue build you laid out for next year, but will margin drag the way it is here in Q4 before revenue actually steps up? Any color would help.
Ed McGowan
executiveYeah, sure; so just to paint a bit more of a picture for you, typically when you undertake these large build-outs, your colocation cost is going to arrive before revenue does. Another factor is that long-term colo commitments come with price escalators, so you have to account for the straight-line effect of those. What happens as a result is you take a non-cash charge, because you are in fact booking a cost that is higher than what you pay for colocation. Our estimate is that this will create roughly 1 or 2 points of margin pressure next year. That said, I think margins will probably bottom out in the first half, maybe a point or so below where we are now, and we should be running in the mid-20s, bearing in mind that you have got this 1 to 2 point non-cash drag on margin that more or less resolves itself over time. Another thing is that with some of these build-outs, there are additional costs tied to what we call network build and support. We incur some one-time costs for items such as freight, small equipment, consulting, contractors who go and build, and things of that nature; that can create some near-term pressure as well. But as the slide we provide to you shows, what you will see happen here is margins more or less bottom out in the first half of next year, begin to improve, and the place where you will really see a ton of leverage is on the EBITDA margin. That should expand quite significantly through 2028.
Jackson Ader
analystUnderstood. All right. Great, that's genuinely useful. So then, Dr. Tom, the CapEx per megawatt comes out higher than the neoclouds, but that's actually a positive. I just want a brief explanation of why. Thanks.
Tom Leighton
executiveExcellent question. At Akamai, CapEx per megawatt runs higher since fewer megawatts go into our CapEx. When the denominator shrinks because you're drawing less power, the CapEx-to-megawatt ratio grows. The less power we consume and the more efficient we get, the larger our CapEx per megawatt becomes; in fact, imagine we could run our servers on zero power — then CapEx per megawatt would be infinite. The problem people run into when they model on a per-megawatt basis is that you assume, "Oh, that figure ought to be low." Actually, you want it high, since that signals you're consuming less power for your infrastructure. That's a major advantage for us; and revenue per megawatt is what counts, not just CapEx, and our revenue per megawatt is very strong; but that helps —
Ed McGowan
executiveOne more point worth mentioning: revenue per megawatt tends to run higher on CPU deals compared with GPU deals. Take two otherwise identical deals — say one at $15 million a megawatt, where a single megawatt runs you roughly $4 million in colocation. That gives you $4 million of depreciation plus $1 million in other costs, which lands you at a 40% margin. Now, on a CPU deal you might be getting $25 million per megawatt, again with $4 million in colo costs, $10 million of depreciation, and $1 million of other costs — and that is also a 40% margin. So keep in mind that you do get more revenue per megawatt out of CPUs than out of GPUs.
Tom Leighton
executiveCPUs are more power efficient, so you're in a better position.
Vijay Homan
analystHey, guys. Vijay Homan here, filling in for Jeff. This is my first one. It seems like the drop in delivery and other cloud apps slowed down somewhat this quarter. Do you have any update on delivery pricing or bit growth, plus what you expect for those two things and how they'll change over the coming 6 to 12 months?
Ed McGowan
executiveYeah, good question. On pricing, really nothing. I'd say we're holding pricing better right now than we have in a long stretch of time. Actually, looking back over the last four quarters, delivery in cloud apps is fairly stable, right? Each quarter it stays within plus or minus $1 million or $2 million, and that's typically seasonality. You get some seasonality in there, maybe a big renewal shows up. But broadly speaking, it's kind of flattish. Looking ahead, I think we're in that stable zone, down a point or two or three or whatever; but it's stabilizing, pricing is improving a little, and we're being quite selective going forward here, particularly in our big media business.
Vijay Homan
analystGot it, that's helpful. One quick follow-up on the LayerX acquisition, which closed a couple of months ago now. Could you give a brief update on how the integration is going, and maybe share your thinking on the revenue potential over the next couple of years or so? Thanks.
Tom Leighton
executiveThings are going really well on the integration front, and sales are already substantial, with some major enterprises on board, among them a top U.S. satellite and streaming TV company that's using it. I'd say the growth potential there is very strong. LayerX's role is to prevent staff from putting confidential company information and data into prompts while they work with AI, along with other safeguards. In other words, it keeps your workforce protected during AI use. Response from the analyst community has been very strong, and in the customer feedback captured in their latest report, we scored quite well—4.8 out of five on customer satisfaction. Naturally, it's still early, but integration has gone smoothly and results have been solid up to now.
Abhishek Murli
analystHello, Abhishek Murli here, filling in for Sanjit Singh. Appreciate you taking my question. This one may be for Ed. In the past, you've mentioned a readiness to go below investment grade, and my understanding is that part of this is somewhat mechanical as those deals ramp, and the EBITDA disclosure helps with that. I'm curious how you're viewing that dynamic moving forward, particularly given the recent increase in scrutiny around financing, and anything you can share regarding cost of debt or cost of equity.
Ed McGowan
executiveYeah, sure. You cut out a bit there. I didn't quite catch exactly what you said. You mentioned we made a comment about investment grade. Could you just repeat that part?
Abhishek Murli
analystYeah, absolutely. Earlier you mentioned a willingness to possibly not — or to drop below investment grade — so I'm just wondering how you're looking at that, given the broader scrutiny on financing, and whether you folks are still open to that possibility.
Ed McGowan
executiveYeah, well, just so we're clear, our aim would be to remain investment grade for as long as we certainly can. If a moment arrived when you had to raise additional capital in order to go below investment grade — which we don't believe we're in that position — but keeping investment grade is helpful. It's actually for our colo investments that we make with our colo providers. Poss Since that's the key measure, and considering where our leverage ratio stands today, the EBITDA we'll bring in over the coming two years alone will greatly expand our borrowing capacity while keeping us within investment grade territory. That leaves us with a huge degree of flexibility, so I'd say we're in excellent shape.
Abhishek Murli
analystThat's really useful. Thanks.
James Fish
analystHi, everyone. I wanted to revisit the earlier question from Mr. DiFucci about confidence, because I understand this is a new line of business for you, and on the last print you did mention a deceleration in compute; but let's be honest, we've essentially missed compute or come in only in line nearly every quarter since you've had this business, and now you're issuing a target that is far above what any of us anticipated. What's behind that confidence, or why do you feel compelled at this point to put it out when, frankly, we haven't really seen you hit your compute numbers yet? Ed, could you walk us through what has to happen to get to the $1.1 billion rather than the $1.3 billion next year, and just help us help you.
Ed McGowan
executiveYeah, sure. I mean, I won't go back over the past, so let's just focus on moving forward. Looking at the range we've given, obviously, a few things need consideration. First, is the contract signed so we can hit those numbers? The answer there is yes. Second, do you have all the power that you need? The answer is yes. Do we have all the equipment that we need? Most of that is already in hand; what's left comes down largely to timing—we'd rather not hold onto it too long and bear the full cost of everything in the meantime. So we're able to secure the whole amount, and there's a pricing mechanism in place that lets us adjust for it. At this stage, all that's left is that we'll have three of these deals ramped up and producing revenue. Equipment arrives on a set schedule, we install it in the data center, and once it's been tested and is operational and made available, the customer can begin using it and we begin billing. From a high-level view of the remaining work, there isn't much risk. Could timing issues arise that push something past our schedule? That's basically what occurred — we had assumed we'd be in the data center on a certain date and that all our equipment would arrive on a certain date, and things slipped a couple weeks. That happens, right? What we've done now is give you a wide range to show that, look, if we were to experience some delays relative to what we have as a fairly conservative view of the world, yeah, we'd be down close to the lower end. Should our delivery timing improve—or even get modestly better—we'd be positioned to land at the top end of that range. On top of that, the range we've given doesn't bake in much added productivity, since factoring that in would introduce considerably more risk, correct? Were I to fold in big contracts that haven't been inked yet, the honest truth is that when those actually get signed isn't something I control. That kind of approach carries greater risk. What this range rests on is agreements already signed, with the space locked in and set up, and our supply chain ready to go.
James Fish
analystThe thing people keep asking us is — clearly, we've got these big contracts signed now, more than $14 billion, so congratulations on hitting that figure, as far as what you all need to go deliver on. Given the current climate, how willing are you to sign another big deal before you've executed on these ones at this point?
Ed McGowan
executiveYeah, Jim, I'm glad you brought that up, since in my prepared remarks I did mention, regarding that $7 billion figure we're discussing, when we said we expect to exceed it in 2028, I noted that we don't expect to take any of those further tranches from the Anthropic relationship. That's not to say we're unwilling or anything of the sort. We simply figure that, first, we need to deliver on what's already in front of us, and second, there's an incentive on their side as our stock climbs with the warrants attached, so we just don't assume they'll be aggressively pursuing those orders just yet. I think once we've executed through that, the willingness on our part to move and grow more quickly is there; we do intend to have capacity — both capital and data center space — for bigger deals, meaning something in the several hundred millions and that kind of thing. But when it comes to the multi-billion dollar deals, that isn't in our forecast or our plans. Let's get through the execution here over next year, and then we'll discuss it more at that point.
Param Singh
analystYeah. Hi. Thanks. Thanks for taking my question. First off, I was hoping to get a feel for how security grew; it looked a little soft this quarter. How do you expect that to play out over the next year or two, as you discuss the cross-sell more with API Security and the other modules you brought up earlier on the call? Then I had a follow-up question. Thank you.
Ed McGowan
executiveYeah. I'm not sure. Were you referring to the quarter we just reported? I believe we came in a bit ahead of where you folks had it. Or are you talking more about the guidance?
Param Singh
analystI suppose I was running a bit ahead of the Street there on the security portfolio for this quarter, but I'd really like to focus on the coming two years and how that feeds into your $7 billion guide, since we know the $2.2 billion coming from CIS, so a bit of detail on that would help.
Ed McGowan
executiveYes, absolutely. We won't provide guidance on each individual component, particularly for 2027. That will come on the February call. And for 2028, certainly, we'll give you guidance on that as well. But I think you can assume more or less normal business as usual. We've been fairly consistent — if you look at sequential growth and I adjust for some of those unusual licenses from last year, we've been adding 12 to 13-ish, perhaps slightly more, of sequential additional security growth each quarter. So it keeps growing at a very healthy rate. We have a long way to go with our API Security, and obviously LayerX is brand new. Guardicore is still growing at a very healthy rate and has relatively low percentage penetration in the base compared to some of our other products; we're adding new customers at a faster rate; we've made an investment in adding customers. So, we feel pretty good that security will keep growing, and we'll give you more guidance on that in February regarding exactly how fast next year.
Param Singh
analystGot it. For my follow-up, I don't want to keep harping on this, but I understand your point about why CapEx per megawatt runs higher, and you're also mentioning higher revenue—still, let's set the middle piece aside, okay? If we look at revenue divided by CapEx, since you're normalizing for megawatt efficiency—in that case, I mean, it still looks fairly thin, right? On a revenue-per-megawatt basis, it appears lower than many other peer hyperscalers, and clearly consistent with some of the neoclouds' older contracts. So the dynamics behind this deal are very hard for me to grasp. If you could just break it down from a revenue-per-CapEx angle, that would help me get a better handle on this. Thank you.
Ed McGowan
executiveYeah; so you need to be certain you're making an apples-to-apples comparison. If you're comparing, say, the GPU rental business and what folks receive on short-term contracts, plus how CapEx behaves relative to a megawatt of power for a GPU in that setting, against a seven-year commitment for CPU, those two can't be compared. They are extremely, extremely different. What I believe I heard you say is that our revenue per megawatt resembles what you see at hyperscalers. I think that's what you said. Their mix there is mostly, primarily CPU. I don't know their CapEx offhand from that angle. But once more, we're showing you the economics for the absolute best pricing available when you have significant volume over a significant term. If I approach the big three hyperscalers today and want a one-year deal with no commit, versus a three-year or a five-year deal with a large commit, I will get dramatically better pricing for the longer-term, higher-commit arrangement; it's difficult for me to answer when comparing against these other companies, because I'm not certain what you all are examining regarding what the dynamics are. Are you referring to a short-term contract, a long-term contract, and so on? I simply don't know their figures offhand. But we're very happy with the returns this business is generating, and you can see we're headed into a stretch of significant revenue growth and significant profit growth.
Rishi Jaluria
analystOh, great. Thanks a lot for the question. Yeah, so I guess I'd like to ask something more about the technology infrastructure side. As you keep expanding this fleet, not only for those CIS deals you're signing, but I think for demand you're anticipating down the road — Tom, you touched on this somewhat in your prepared remarks — but obviously there are plenty of other potential use cases out there to pursue, physical AI, presumably sovereign AI, particularly given the sort of capability growth we're seeing around open-weight models. Could you maybe take us through what you're doing to make sure the fleet you're building is completely fungible, so that you can shift in real time toward however adoption trends in AI tend to take shape, especially given that leadership and models tend to change? Maybe your thoughts on why you feel Akamai is uniquely positioned to benefit from both of those trends. Thank you.
Tom Leighton
executiveYeah. Well, actually, we've signed some very large physical AI deals, quite a few physical AI contracts already. Sovereign AI also makes a lot of sense for us given our distributed platform; we do have diversity in the hardware across our platform, in the servers. But as we discussed, we put the emphasis on the CPUs. They're extremely flexible. And increasingly, for the kinds of enterprise applications that our customer base is running, CPUs perform very well. Again, we aren't pursuing training the next frontier model. What interests us is post-training of, say, open source models on our platform, now with our Akamai Inference capability, and having agents on our platform with Agent Sandbox; and we'll assign that to whichever hardware is most efficient and capable of handling the particular workload, and we take that responsibility off our customers' hands. I think we're really ideally positioned with our distributed platform to get these workloads onto the right kind of server that makes it efficient. We can get it close to the robot, if that's what it's being used for, the sensor, the alert sensor, if that's what it's being used for, or the user, if you're trying to maintain some kind of dialogue or chat with the user or some video with the user, and their proximity matters for latency and scale. And we're really unique at that; and that's why even the hyperscalers use our compute platform today for the latency-sensitive applications. I think, yeah, and you compare it to the neoclouds, it's an entirely different capability; and then, of course, with Akamai packaged also with the security you need, especially important for AI, and the delivery. A lot of these applications involve some form of video with the user, or video with a robot or a car, and in those cases you need the kind of scalability that our edge platform provides.
Rishi Jaluria
analystThat's quite useful. Thanks.
Peter Levine
analystGood. Appreciate you fielding my questions. Perhaps only two brief follow-ups. Regarding that 13% IRR, my assumption is it's a blended figure, clearly, based on your slide deck, but considering how large Anthropic is, would it be reasonable to think Anthropic got more favorable pricing than your other customers, and so produces a lower return? Just help us a bit, take us through how we ought to think about the spread between some of your returns and those of these bigger customers.
Ed McGowan
executiveYes, I believe you captured it nicely, that from a pricing standpoint they will receive a superior return since they are, first, prepared to sign on for seven years, and second, they are committing substantially greater volume. So that would be the lowest kind of return you'd see. Customers who don't commit to such a term or don't commit to that volume will end up with higher IRRs.
Peter Levine
analystPerhaps just following up on what John DiFucci asked, but I suppose leaning more toward the political angle. Would you see this as a boost, I suppose, for your CIS business, if we assume you're able to handle additional AI workloads inside the colocation footprint you already have? Could you just take us through what happened after the midterms? Is it a boost? Is it a drag? Or does it simply not really alter how you view this business?
Tom Leighton
executiveWell, as far as the business goes, it doesn't hurt us. In some ways, we're comparatively better off than competitors, who I think do face serious difficulties. Of course, the market became quite frightened, as you know, and we were pulled down along with it, I think unjustly, since the problems that frightened the market truly don't apply to Akamai. As I've said, we're in a distinctive and excellent situation to secure the colo and power we require, which comes in numerous locations with far smaller footprints. As Ed mentioned, we already hold permits for everything we need through 2028 for all the deals we've discussed and beyond; so that's not a problem. Now, if it does push prices upward, push costs upward, well, once again, we're under contract, so that won't affect us. It could potentially let us increase prices, and that might be a comparative edge, since the other players will potentially face difficulties with these matters; with our distributed platform, we're truly in an excellent position not to have those sorts of problems.
Roger Boyd
analystGreat. Appreciate you fitting in the question. Dr. Tom, you raised a key point — the CIS business is 90% compute-driven, and you're not selling bare metal, so much of that compute can be shifted around the rest of the platform. As you look at the capacity investment required to scale security and delivery over the coming couple of years, do you see a point where you'd be less inclined to take on new delivery and security business, given the opportunity and margin uplift on the compute side? Thank you.
Tom Leighton
executiveWell, clearly security is a very good business for us. Roughly $2.5 billion, up 10% in constant currency just now; we keep investing in and growing that business. Delivery, as we've discussed, we aren't investing in the same way there. Of course, we want the platform to support all the new protocols, be very reliable, and perform well. But we won't pursue the very low-end deals on pricing, or those with huge spikes, where it simply doesn't make economic sense. We're far better off putting that CapEx into our CIS business, much better returns; but I wouldn't equate security with delivery. Security is a very strong business, and it doesn't require a huge amount of CapEx.
Roger Boyd
analystNice color. Ed, I just want to close the loop on that $7 billion target. If you figure security and delivery keep following a steady trend across the next few years, it seems like you could hit that target simply by executing on the $2.2 billion runway backlog you have in compute today. Is that the correct way to look at it, and is what that implies about the other segments broadly right? Whatever color you can offer on the delivery and security assumptions behind that guide would help. Thanks.
Ed McGowan
executiveYeah. No, I think that is the right way to think about it, that based on what we have signed today it is very achievable, and there is upside to that; now, keep in mind, 2028 is going to be a ramp year, so obviously not all of the $2.2 billion is in for the full year. But yes, you are thinking about it the right way, that based on what we have signed it is a very achievable number; we are not expecting massive big deals; that is why I made that comment about not assuming there is any additional Anthropic revenue in there, which obviously would change the trajectory quite a bit. So, you are interpreting it the correct way.
Rudy Kessinger
analystHi folks, thanks a lot for answering my question, and I really appreciate the added detail you provided. So thinking about what's coming, when you look at the big deals in your pipeline and keep in mind that Anthropic secured better pricing, which is likely dragging down that 13% after-tax IRR somewhat — if you consider your pipeline and the large deals currently under negotiation, would those deals on average carry a higher IRR at a comparable level? Would it be a few points above? Would it be significantly above? I'm just hoping to understand what upcoming deals could look like from that angle. Thanks.
Ed McGowan
executiveYeah, that varies. The great part is that we have the freedom to be selective as well. So we decline transactions we feel fail to clear our profit hurdle or return on capital. Still, I'd note that in many instances those are considerably above what we're looking at here today. Then again, nobody is out there deploying $11 billion at once, which makes this quite unusual. A player putting in a couple hundred million would earn a far better IRR than that.
Operator
operatorThat wraps up both the Q&A portion and today's conference call. We appreciate your participation in today's presentation. You're welcome to disconnect at this time.
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.