NextFin News - Akamai Technologies shares jumped as much as 20% in after-hours trading on Thursday after the company announced an $11.6 billion, seven-year cloud computing agreement with Anthropic, the largest contract in its history and a direct bet that AI inference will migrate from centralized data centers to the distributed edge. The deal carries an expansion path worth up to an additional $9 billion - a total potential commitment of roughly $20 billion - and comes with a warrant giving Anthropic up to 5% of Akamai's common stock, tying the AI lab's economics to the very network it will run on.
The Deal: $11.6 Billion Locked, With a Path to $20 Billion
Under the agreement announced September 24, 2026, Anthropic will use Akamai Cloud's distributed infrastructure and software to support its accelerating CPU workload demands. The commitment adds to more than $2.8 billion in multi-year Cloud Infrastructure Services (CIS) contracts Akamai has announced this year, turning a business line that management only recently began breaking out into the fastest-growing part of the company.
The structure is unusual for a cloud contract. Alongside the service commitment, Akamai issued Anthropic a warrant to purchase non-voting convertible Series B Preferred Stock representing 7.7 million shares on an as-converted basis - up to approximately 5% of Akamai's outstanding common stock - at an exercise price of $111.33 per share. Roughly 2% of the company's common stock is expected to vest with the $11.6 billion commitment announced Thursday; the remaining approximately 3% vests only if Anthropic expands the relationship by up to an additional $9 billion within the seven-year term, with each additional $3 billion of cloud services purchases triggering roughly 1% of common stock.
The warrant converts a vendor relationship into an equity partnership. Anthropic now has a direct financial incentive to route more workloads through Akamai and to hit the expansion tranches that unlock the remaining equity. For Akamai's existing shareholders, the trade-off is explicit: contracted revenue visibility in exchange for up to 5% dilution.
The market read the announcement as a transformation event rather than a routine contract win. Akamai closed the regular session at $110.41, down 6.78% on a day when the broader market was mixed, then surged to $129.25 in after-hours trading - a gain of $18.84, or 17.06% - with intraday after-hours buying pushing the move as high as 20%. Even after the rally, the shares sit below their 52-week high of $165.45 and below their all-time closing high of $161.14, both reached in May.
Why CPU Workloads at the Edge Is the Real Story
The most important word in the announcement is "CPU." The hyperscaler AI buildout of 2024 and 2025 was a GPU story: training frontier models requires tens of thousands of accelerators packed into specialized data centers, and that work is not moving to the edge. Inference is different - and agentic AI makes it structurally different.
Every agent action - a search, a tool call, a code edit, a customer-service turn - generates inference requests that must return quickly to feel responsive. Running those requests on infrastructure close to the user cuts latency and, just as important, reduces the cost of moving data back and forth to centralized regions. Akamai's pitch rests on a physical fact: it operates infrastructure in roughly 4,300 locations across 700 cities in 130 countries. In May, CEO Tom Leighton put the logic plainly in an interview.
"We operate the world's most distributed platform, and we have our infrastructure in 4,300 places, 700 cities in 130 countries, and we've used that for delivering content and for providing security to intercept all the attacks, and now we're using it to support AI so our customers, agents and AI apps can run right near their users, and the data provide a much faster experience."
That quote predates this deal, but it is the thesis the $11.6 billion validates. Anthropic has not put all its compute with one provider. Over the past year it has assembled a deliberately diversified stack: Amazon committed $25 billion with up to $20 billion more for Trainium-based capacity under Project Rainier, Google committed up to $40 billion in cash and TPU compute, and Microsoft committed $13 billion. Anthropic has also signed with CoreWeave for Nvidia-based capacity. Against that backdrop, the Akamai deal is not a training contract; it is the inference-and-agentic layer of a multi-cloud hedge.
The scale of the re-rating math is what moved the stock. Akamai's Cloud Infrastructure Services generated $99 million in revenue in the second quarter of fiscal 2026, up 39% year over year - fast growth, but only about 9% of the company's $1.1 billion quarterly revenue. Full-year 2026 revenue is guided to $4.445 billion to $4.53 billion. Spread evenly, $11.6 billion over seven years is roughly $1.66 billion a year - equivalent to more than a third of Akamai's current annual revenue base from a single customer at full ramp. A contract that size does not just add growth; it changes what kind of company Akamai is.
The Warrant, the Dilution, and the Alignment
The equity component deserves its own read. A plain cloud contract aligns a customer only as long as prices stay competitive. A warrant that vests on expansion aligns the customer with the stock itself. If Anthropic expands to the full $20 billion, Akamai issues the full 5%; if the relationship stalls at $11.6 billion, Anthropic keeps roughly 2% and Akamai keeps the other 3% of equity off the table. The vesting is back-loaded and milestone-driven, which means the dilution is not free - it is purchased with incremental revenue.
There is a cost to this structure that the rally glossed over. Seven-year AI infrastructure contracts signed in 2026 were negotiated in a seller's market for compute. If AI demand cools, or if Anthropic's own revenue growth slows, the expansion tranches simply do not vest. The base $11.6 billion is committed, but the headline "$20 billion potential" is conditional, and the market should value the two tranches differently.
There is also a cash-cost side. To deliver this capacity, Akamai is spending heavily. In the second quarter, capital expenditures were $347 million, or 32% of revenue, and management has guided full-year 2026 capex to roughly 40% of revenue, with third-quarter capex projected at $475 million to $525 million, or 43% to 46% of that quarter's revenue. That level of investment compresses free cash flow and explains why the company paused share buybacks: it repurchased about 3 million shares for $410 million in the second quarter, and roughly 5 million shares for $616 million year to date, before CFO Ed McGowan said the company was "temporarily pausing share repurchases to reallocate capital to support our high growth CIS pipeline." The bull case requires that this spending converts into contracted revenue faster than it drains cash.
Cyclical Wave Meets Structural Shift
Is this deal a cyclical AI capex peak or a structural repositioning? The answer is both, and separating them matters for valuation.
The cyclical leg is real. The AI infrastructure buildout of 2024-2026 is the largest single-purpose private-sector capital deployment in modern U.S. history, and cycles this large do not go straight up. If AI monetization lags that spending - if token prices keep falling and enterprise adoption proves slower than the capacity build - the entire complex enters a digestion phase. In that scenario, Akamai's expansion tranches stall, its 40%-of-revenue capex becomes a drag instead of an investment, and the stock reverts toward legacy CDN multiples.
But the structural leg is the stronger force here. Three pieces of evidence point to a regime shift rather than a cycle top. First, the workload is inference and agentic AI, which grows with usage rather than with training runs - every new agent user generates recurring, distributed compute demand. Second, once an AI application is architected to run inference at the edge, migrating it back to centralized regions means rewriting latency assumptions; the switching cost is architectural, not just commercial. Third, Akamai's legacy delivery business is already in structural decline - delivery and other cloud applications revenue fell 6% year over year in the second quarter - so the company is not adding AI on top of a healthy core; it is replacing a shrinking core, which makes the transition existential rather than optional.
The structural call: the migration of AI inference to distributed infrastructure is a regime shift that will not revert on its own. The cyclical call: the pace of contract announcements and the multiple investors assign to them will fluctuate with AI sentiment. The trade is long the structural shift, but the entry point depends on the cycle.
The Counter-Thesis: Margin, Competition, and a $93 Price Target
The strongest case against the rally is not that the contract is fake - it is signed - but that it arrives too late and at too low a margin to justify the re-rating. CPU-based inference is less scarce and less lucrative than GPU training capacity. The hyperscalers are not standing still: Amazon, Google and Microsoft are all building their own edge inference offerings, and they can bundle compute with the model platforms and developer tools Akamai does not own. Akamai's own capex guidance - roughly 40% of revenue in 2026 - means free cash flow will be under pressure precisely when investors are being asked to pay an AI-infrastructure multiple.
Wall Street's skepticism is quantifiable. As of mid-September, Goldman Sachs maintained a Sell rating on Akamai with a $93 price target, set on September 15, 2026 - more than 15% below the pre-announcement close. Citigroup cut its target to $122 from $160 on September 4 while keeping a Neutral rating. The consensus view, before Thursday, was that Akamai was a Hold with an average price target in the low-to-mid $140s. The bull argument that this is "no longer a legacy tech company" is exactly the claim the skeptics deny.
The answer to the counter-thesis rests on the contract's structure and the customer's behavior. The $11.6 billion is a committed contractual obligation, not a forecast, and it arrives on top of a CIS line already growing 39% to 40% year over year. More importantly, Anthropic's multi-cloud pattern - spreading tens of billions across Amazon, Google, Microsoft and CoreWeave - shows a deliberate strategy of avoiding single-provider lock-in. That diversification is Akamai's opening: it does not need to beat the hyperscalers at their own game, only to win the slice of inference workloads that hyperscalers are structurally worst suited to serve. The falsifying signal is specific: if Cloud Infrastructure Services revenue growth falls below 30% year over year for two consecutive quarters after the fourth quarter of 2026, the structural re-rating thesis is wrong and the cyclical-digestion view wins.
What Comes Next: Scenarios and Signals
For investors, the deal splits into three time horizons with different implications.
In the short term - the next two quarters - the stock will trade on sentiment and on whether the after-hours gap holds. The key data point is third-quarter and fourth-quarter CIS revenue, where management has guided the full-year line to grow at least 50% in constant currency. A clean beat would confirm the deal is already ramping; a miss would suggest the revenue recognition lag is longer than the market hopes.
Over the medium term - 2027 and 2028 - the story is fundamentals and cash flow. The base case is that the $11.6 billion ramps steadily, CIS becomes 20% to 30% of total revenue, and Akamai earns an AI-infrastructure multiple closer to its edge-computing peers than to legacy CDN names. The upside case is that Anthropic hits the expansion tranches, the full 5% warrant vests, the relationship reaches the $20 billion potential, and Akamai is re-rated as a core AI infrastructure platform. The downside case is that AI capex enters a digestion phase, the expansion tranches do not vest, capex stays near 40% of revenue while growth disappoints, and the multiple compresses back toward the low-teens forward earnings range that skeptics like Goldman Sachs have assigned.
In the long term - the seven-year life of the contract - the question is whether distributed inference becomes the default architecture for agentic AI. If it does, Akamai's 4,300-location footprint becomes a moat that hyperscalers cannot replicate without rebuilding their networks from the edge inward. If agentic AI instead consolidates into a few centralized superclusters, the edge thesis weakens and the deal becomes an expensive capacity bet.
The signals to watch are concrete: quarterly CIS revenue growth against the 50% full-year guide; capex as a percentage of revenue and whether free cash flow recovers after 2027; the pace of Anthropic's own usage growth, which drives the expansion tranches; and any movement in analyst targets from the hold-and-sell camp. The warrant vesting schedule - roughly 1% of common per additional $3 billion of purchases - provides a public milestone tracker for the relationship's expansion.
One line captures the stakes. This is not primarily a story about $11.6 billion in contracted revenue; it is a story about who owns the last mile of AI, and Akamai has just bought a claim on it - with Anthropic's own money.
Explore more exclusive insights at nextfin.ai.
