Have you ever watched a stock sit quietly for months, then suddenly wake up because one contract changed the whole story? That is pretty much what happened with Akamai this week. Shares jumped hard after the company signed a long computing agreement with Anthropic, then gave some of those gains back as traders started doing the math. I have seen this pattern before. The headline is loud. The balance sheet work comes later.
Why Akamai Shares Jumped After The Anthropic Pact
The agreement is large enough to force a rethink. Akamai will supply dedicated computing capacity and managed support over seven years in a deal valued at about $11.6 billion. There is also room to expand the hardware terms and lift the total toward $20 billion. Anthropic is taking a stake in Akamai as part of the arrangement. That detail matters. It is not just a purchase order. It is a tighter commercial relationship.
The market reaction was fast. Shares briefly climbed as much as 16.4% to $128.46 before settling near an 8% gain. For the year, the stock is still up about 36%, though almost all of that move arrived early. In my experience, a midyear surge after a sleepy stretch often says more about expectations than about yesterday’s price.
Bookings momentum in cloud infrastructure is starting to look like proof that Akamai’s growth profile is changing in a meaningful way.
One more point that is easy to miss. This capacity is aimed at CPUs, not the GPU clusters that dominate most AI headlines. Networks of CPUs are often better suited to coordinated work, the kind of multi-step activity people now call agentic AI. That does not make GPUs irrelevant. It does mean the market may have undercounted a second hardware path.
What The Contract Actually Covers
Akamai is not handing over a generic cloud login and walking away. The company is providing dedicated capacity plus related managed services. The master services agreement is split into projects. This latest commitment covers additional phases of the Anthropic relationship rather than a one-off burst of demand.
That structure has two effects. First, revenue should arrive over time instead of in a single spike. Second, Akamai has to build the capacity itself. Management pointed to about $5.5 billion in capital spending tied to the work. Year to date, total cloud commitments have reached roughly $14.6 billion. That is a lot of future work sitting on the books.
- Seven-year computing and support commitment
- Core value near $11.6 billion, with an option toward $20 billion
- CPU-focused capacity, with room to widen hardware terms
- Customer equity stake alongside the commercial contract
- About $5.5 billion of expected build-out spending
I keep coming back to the equity piece. When a customer also becomes a shareholder, incentives line up in a different way. Delivery quality, pricing discipline, and long-term capacity planning stop being abstract. They become shared problems.
Why CPU Capacity Is Suddenly Interesting
Most AI conversations still orbit graphics processors. Fair enough. Training giant models eats GPU hours. But once software starts acting across tools, files, calendars, sensors, and live video, the workload changes. Coordination, memory movement, and steady inference can lean on large CPU fleets.
Think of it like a kitchen. The specialty oven matters. So does the prep table. Agentic systems need both. Conversational voice agents, robotics, physical AI, real-time virtual worlds, and live video intelligence all showed up in management’s broader cloud pitch. Those are not purely training jobs. They are always-on systems.
Perhaps the most interesting aspect is timing. If frontier labs keep stacking agent features on top of base models, CPU demand may rise even when GPU supply gets tighter or more expensive. That is a second engine, not a replacement engine.
How Analysts Are Recasting The Growth Story
Street notes after the announcement were unusually aligned on one idea: Akamai is drifting from a slower-value profile toward a faster compute story. One team used the midpoint of management commentary and said the deal could add $150 million to $300 million of revenue in 2027. That would lift modeled growth from about 12.5% toward roughly 16%. The company reported only about 5% revenue growth last year. That gap is the whole debate.
The size of this win changes the financial profile. Cloud infrastructure could overtake security as the larger growth engine later this decade if the current booking pace holds.
Another note put a buy case around dedicated capacity for accelerating CPU workloads and a wider vision for Akamai Cloud. Use cases listed by the company were not vague slogans. Voice agents. Robotics. Simulation rendering. Real-time video intelligence. Those examples make the contract feel less like a one-customer anomaly and more like a product direction.
A separate analysis tried to translate dollars into power. Using an industry-style average near $22 million of annual revenue per megawatt, a contracted run-rate around $1.7 billion implied roughly 77 megawatts. Akamai has not published a deal-specific power figure, so that number is an estimate, not a filing. Still, it helps frame the build. Capex near $5.5 billion would land around $70 million per megawatt in that framework, including a large memory pre-buy.
| Item | Figure Discussed | Why It Matters |
| Contract value | $11.6B, option toward $20B | Sets multi-year revenue visibility |
| 2027 revenue add | $150M to $300M | Could lift growth into the mid-teens |
| YTD cloud commitments | $14.6B | Shows bookings are broader than one deal |
| Expected capex | $5.5B | Cash and execution become the next test |
| Implied capacity | ~77 MW estimate | Helps size the physical build |
Price targets clustered in a familiar band after the news: mid-$150s to the mid-$180s, with overweight and outperform language across several desks. I would not treat any single target as destiny. Targets move when the next quarter either confirms the bookings story or shows slippage in delivery.
Margins, Capex, And The Unsexy Part Of The Win
Big contracts look glamorous until the invoices for concrete, power, memory, and networking show up. Akamai has to stand up capacity to fulfill the work. That $5.5 billion build is the trade-off. Spend now, harvest later.
On the brighter side, analysts looking at the $14.6 billion of cloud commitments signed this year have talked about healthy economics once those deals reach a full revenue run-rate. One estimate put EBIT margins near 30% at maturity for that book of business. If that lands, the stock conversation shifts from “nice booking” to “durable profit pool.”
Still, maturity is the key word. Construction delays, power constraints, component shortages, or weaker utilization can push that margin story to the right. I have found that markets celebrate the signature and then get impatient about the shovel.
- Book the multi-year capacity commitment.
- Fund and build dedicated infrastructure.
- Bring projects online in stages.
- Convert contracted dollars into recurring revenue.
- Defend margins once utilization rises.
That sequence is simple on a slide and messy in real life. Power availability alone can decide whether a 77-megawatt estimate is a planning tool or a headache.
Does One Customer Change The Whole Company?
Not by itself. Concentration risk is real. A seven-year pact with a frontier lab is impressive. It is also a reminder that a handful of AI buyers now shape demand for an entire class of infrastructure. If Anthropic’s own product plans shift, Akamai feels it.
The counterargument is that the partnership does not lock Akamai into a single lab. Additional CPU or GPU wins remain possible. Management’s broader cloud examples suggest the company wants a portfolio of workloads, not a monument to one logo. I think that is the healthier reading. One jumbo deal can re-rate a stock. A series of jumbo deals creates a new identity.
Security remains an important franchise. The new claim is not that security disappears. The claim is that compute may grow fast enough to overtake it as the narrative driver by late decade if bookings stay hot. That is a high bar. It is also why Friday’s move was so sharp.
How Investors Might Think About The Stock From Here
After an 8% to 16% day, the first question is always the same. Did the multiple already swallow the news? Maybe part of it. Maybe not the whole path to 2028. The honest answer depends on three things: delivery speed, incremental wins, and how expensive the build becomes.
If 2027 revenue really steps up by a couple hundred million and other cloud deals keep landing, growth in the mid-teens starts to look less like a stretch. If capex overshoots or utilization lags, the same contract becomes a cash-flow story instead of a multiple story. Both outcomes are possible. Markets hate that sentence. It is still true.
A contract this size can turn a value-looking infrastructure name into a growth debate overnight. The debate only stays won if the data centers actually come online.
I would watch four markers. New cloud bookings after this announcement. Commentary on power and construction timelines. Early 2027 revenue contribution. And whether memory and networking costs stay inside the implied build budget. Those are not glamorous. They are the difference between a one-day spike and a multi-year rerating.
A Practical Read On Risk Without The Drama
Every large AI infrastructure deal carries familiar risks. Customer concentration. Technology mix. Power. Capex timing. Competitive pricing from much larger clouds. Akamai is not the biggest player in this arena. That can be a feature when a lab wants dedicated capacity and closer service. It can be a bug if hyperscalers decide to match terms aggressively.
There is also narrative risk. The market is hungry for any stock that looks like a second-wave AI winner. Hunger creates premium prices. Premium prices punish small misses. If you buy the acceleration thesis, you are also buying the expectation that execution stays clean.
- Delivery and power risk on the $5.5 billion build
- Revenue timing that stretches across many years
- Customer concentration around frontier AI demand
- Competition from larger cloud platforms
- Valuation sensitivity after a sharp one-day move
None of that makes the deal a bad deal. It makes it a real business. Real businesses have friction. Friday’s tape just decided the friction was worth paying up for, at least for a morning.
The Bigger Shift Hiding Under The Headline
Zoom out and the story is larger than one ticker. AI demand is splitting. Training still loves specialized accelerators. Deployment, agents, simulation, and always-on inference can lean on different silicon and different networking designs. Companies that can sell dedicated CPU capacity with serious support may find a lane that looked boring two years ago.
Akamai spent years known for content delivery and security. Those businesses did not vanish. What changed is the market’s willingness to believe compute can become a second core. An $11.6 billion agreement is a loud way to make that argument. An expandable $20 billion ceiling is louder.
Will every follow-on deal look like this one? Unlikely. Jumbo contracts are rare by definition. But the direction is the point. Bookings first. Capacity next. Mix shift after that. If those three steps hold, the stock stops being a debate about last year’s 5% growth and becomes a debate about the late-decade mix.
What I Keep Coming Back To
The surge was not mysterious. A long-duration computing contract with a frontier lab, an equity overlay, a path to a larger total, and a CPU angle that fits agent-style work is a lot of news for one morning. The fade from 16% to 8% was not mysterious either. Traders started counting capex and asking when cash turns into earnings.
That is the right argument to have. Not whether the announcement was large. It was. The live question is whether Akamai can turn a stack of cloud commitments into a cleaner growth algorithm without letting the build swallow the upside. I think that is the only frame that lasts past the next session.
If the company keeps landing CPU or GPU work, if 2027 contribution lands in the cited range, and if margins on the committed book get anywhere near the mature estimate, the rerating has room to keep working. If the next updates are mostly construction talk and little new demand, the stock will ask for proof again. Markets are like that. They cheer the signature. Then they wait for the lights to turn on in the hall.
For now, Akamai is no longer just a quiet infrastructure name having a decent year. It is a company with a very public test on its hands: convert one of the year’s loudest computing pacts into durable growth without losing the plot on cost. That test will take longer than a Friday rally. It is also why the rally happened in the first place.
