When Amazon poured its first $1.25 billion into Anthropic back in September 2023, Wall Street shrugged. It looked like a desperate catch-up play, a late swing at Microsoft’s increasingly cozy relationship with OpenAI. Fast forward to February 2026, and that $8 billion total investment has become the most strategically valuable bet in Amazon’s history. The AI startup behind Claude is now closing a $20 billion funding round at a $350 billion valuation, AWS just posted its fastest growth in 13 quarters, and Amazon’s paper gains from the Anthropic stake have already added $9.5 billion to a single quarter’s bottom line.
Not bad for what critics called a “desperation play.”
The Numbers That Shut Up the Skeptics
Amazon’s Q4 2025 earnings told the story in stark terms. AWS generated $35.6 billion in revenue, a 24% year-over-year jump that topped analyst expectations of $34.93 billion. Operating income within the cloud unit hit $12.47 billion with margins expanding to 35%. CEO Andy Jassy called it AWS’s “fastest growth in 13 quarters,” and he wasn’t being modest.
Anthropic is a massive piece of that story. The AI startup’s revenue run rate surpassed $9 billion last year, with a significant chunk flowing directly through AWS infrastructure. Rothschild & Co analyst Alex Haissl estimated that Anthropic added one to two percentage points to AWS growth in recent quarters, with that contribution expected to exceed five points in the back half of 2025. Wedbush’s Scott Devitt went further, telling CNBC that once Claude becomes a default enterprise developer tool, the usage feeds directly into AWS revenue for “many, many years.”
And that $11 billion data center campus in Indiana called Project Rainier? It’s running. Nearly 500,000 Trainium2 custom chips powering Anthropic’s Claude models, with plans to scale beyond one million. Jassy told analysts on the earnings call that Trainium offers 30 to 40% better price performance than comparable GPUs, and combined with Amazon’s Graviton CPU chips, the custom silicon business already exceeds a $10 billion annualized run rate.
Anthropic’s Valuation Explosion
The timeline of Anthropic’s ascent reads like a venture capital fever dream. A $183 billion valuation in September 2025. A $350 billion valuation by January 2026. And now a $20 billion funding round closing this week that doubled its original $10 billion target because investor demand was simply too intense.
According to Bloomberg, the round features checks north of $1 billion each from Coatue Management, Singapore’s sovereign wealth fund GIC, and Iconiq Capital, with the bulk coming from strategic partners Nvidia and Microsoft. That’s right: Microsoft, already deep in the OpenAI trenches, is backing Anthropic too. Satya Nadella’s hedge-every-horse strategy continues unabated.
The investor roster also includes Altimeter Capital, Sequoia Capital, Lightspeed Venture Partners, and Menlo Ventures. Bloomberg additionally reported that Anthropic is working on a tender offer to let employees sell shares at the $350 billion valuation, the kind of liquidity event that keeps top AI researchers from jumping ship to competitors offering public-company stock.
For Amazon specifically, the math is staggering. Its $8 billion investment, originally structured as convertible notes, triggered a $9.5 billion pre-tax gain in Q3 2025 alone when Anthropic’s valuation jumped to $183 billion. At $350 billion, that paper profit has roughly doubled again. Amazon’s minority stake in Anthropic is approaching the kind of valuation inflection that turns strategic investments into generational wealth creation.
Why This Partnership Actually Works
The Amazon-Anthropic relationship isn’t just money. It’s infrastructure lock-in at a scale that competitors struggle to replicate.
AWS serves as Anthropic’s primary cloud provider and primary training partner. Project Rainier represents one of the world’s largest AI compute clusters, purpose-built for Claude workloads. Tens of thousands of companies access Claude models through Amazon Bedrock, AWS’s managed AI service. The European Parliament, Pfizer, and a growing list of enterprise customers have adopted Claude through AWS infrastructure.
That flywheel is exactly what Amazon needs. Enterprise customers sign up for Claude through Bedrock, which means they’re also consuming AWS storage, networking, and compute resources. The more popular Claude becomes, the stickier AWS becomes. It’s the same platform gravity that made AWS dominant in traditional cloud computing, now supercharged with AI demand.
Jassy told analysts that AWS’s backlog hit $244 billion in Q4, up 40% year over year. That backlog includes massive commitments from AI companies, with Anthropic and (now) OpenAI contributing significantly. OpenAI itself recently committed $38 billion in AWS spending, which means Amazon is collecting infrastructure revenue from both leading frontier labs simultaneously.
The Complicated Part: Amazon Is Dating Everyone
Here’s where the narrative gets messy. Amazon is reportedly in talks to invest up to $50 billion in OpenAI, Anthropic’s primary competitor. Jassy and OpenAI CEO Sam Altman have been in direct discussions, according to CNBC, with the term sheet potentially signing in the coming weeks. OpenAI models could even help power Alexa, which currently relies on Claude for many of its complex queries.
That’s not exactly a monogamous relationship. But Amazon seems unbothered by the optics. The company uses upwards of 70 different AI models internally, according to top Alexa executive Daniel Rausch, with the majority of traffic routed to Amazon’s homegrown Nova models anyway. The strategy isn’t loyalty. It’s optionality.
Meanwhile, Anthropic is doing its own diversifying. The company signed a deal with Google Cloud worth tens of billions for one million TPU chips. Then came the Microsoft-Nvidia partnership: $30 billion in Azure compute capacity and a $15 billion joint investment. Anthropic is now the only major AI developer running across all three hyperscale clouds, which gives it negotiating leverage that Amazon probably doesn’t love but absolutely respects.
Reports from The Information in late 2025 suggested the relationship was showing strain, with Amazon’s internal AGI group working on rival models (codenamed Olympus) that aim to compete with Claude directly. The dynamic mirrors what happened between Microsoft and OpenAI: deep partnership combined with quiet competition, held together by financial interdependence and mutual need for infrastructure scale.
The Bigger Picture: AI’s Circular Money Machine
Wall Street analysts have increasingly flagged the circular nature of these AI mega-deals. Amazon invests in Anthropic. Anthropic spends billions on AWS. AWS revenue grows. Amazon reinvests in more AI infrastructure. Nvidia sells chips to everyone, then invests in Anthropic, which buys more Nvidia chips. Microsoft invests in both OpenAI and Anthropic while selling Azure compute to both.
It’s a financial Ouroboros, and the central question is whether the underlying AI demand is real enough to justify the spending. Amazon’s 2025 capital expenditures hit approximately $125 billion, and Jassy signaled that 2026 spending will increase further, with some estimates pointing toward $150 billion. Google just guided $175 to $185 billion in 2026 capex. Meta could spend $115 to $135 billion.
The collective bet across Big Tech now exceeds half a trillion dollars annually. If AI adoption stalls, if enterprise customers don’t convert pilots to production at the expected rate, the entire ecosystem faces a reckoning where everyone’s balance sheet is intertwined with everyone else’s.
Google CEO Sundar Pichai acknowledged the risk explicitly: “No firm is going to be immune” if AI companies overinvest. Nadella echoed the sentiment, calling for the industry to “move beyond any type of zero-sum narrative.” Translation: we’re all in this together, and we’d better all be right.
What Comes Next
Both Anthropic and OpenAI are reportedly preparing for IPOs this year, which would give public markets their first direct stake in the frontier AI race. Anthropic’s $350 billion valuation already approaches the market caps of companies like Netflix and AMD. An IPO would force a level of financial transparency that the private fundraising rounds haven’t required.
For Amazon, the calculus is straightforward even if the execution is complex. AWS needs AI demand to justify its massive infrastructure build. Anthropic is its most important AI customer and the anchor tenant of Project Rainier. The $8 billion investment has produced paper gains that nearly match the original outlay in a single quarter. And Claude’s growing enterprise adoption means more Bedrock usage, more AWS consumption, and a longer competitive moat around the cloud business that generates most of Amazon’s profit.
The risk? Anthropic’s multi-cloud strategy means it’s not exclusively Amazon’s partner anymore. The potential $50 billion OpenAI investment could dilute strategic focus. And the sheer scale of capital deployment across the AI industry leaves every player exposed if the technology’s commercial promise doesn’t materialize fast enough.
But right now, today, the Amazon-Anthropic deal looks like what it always was underneath the initial skepticism: a bet that the company building the safest, most capable AI models would need the world’s largest cloud infrastructure to do it. Amazon had that infrastructure. Anthropic needed it. And $8 billion later, both companies are sitting on one of the most valuable partnerships in tech history.
Call it a desperation play if you want. The scoreboard says otherwise.
