Anthropic Surpasses OpenAI in Business Adoption as $100B AWS Deal Secures 5GW of Compute

Two milestones arrived within hours of each other for Anthropic on June 13, 2026, and together they mark a decisive shift in the competitive landscape of enterprise AI. The Ramp AI Index — which tracks AI software spending across tens of thousands of businesses — showed Anthropic’s Claude at 34.4% of enterprise AI adoption, edging past OpenAI’s 32.3% in what is believed to be the first-ever crossing. Simultaneously, Anthropic’s expanded compute agreement with Amazon formalized the infrastructure backbone required to sustain that growth at a scale previously unimaginable for a five-year-old startup.

The Business Adoption Crossover

The Ramp data measures new adoption — which companies are adding which AI tools to their software spending — rather than depth of existing use. That distinction matters: a separate IDC survey published around the same time shows OpenAI still commanding higher rates of “extensive” use at established enterprise accounts. But the adoption crossover is a leading indicator, reflecting where enterprise IT decision-makers are placing new bets.

Claude Code is the single biggest driver of Anthropic’s enterprise momentum. GitHub data shows Claude Code now responsible for approximately 4% of all public commits on the platform, a figure that has doubled month-over-month. In the increasingly contested market for AI coding assistants — where GitHub Copilot, Cursor, and a growing field of competitors are fighting for developer mindshare — Claude Code’s trajectory is the clearest signal that Anthropic has successfully moved from “impressive demo” to “production infrastructure” status for engineering teams.

More than 1,000 businesses now spend over $1 million annually on Claude, and more than 100,000 customers run Claude via Amazon Bedrock. Anthropic’s run-rate revenue has surpassed $30 billion, a figure that would have seemed fantastical at the company’s $9 billion run rate at the close of 2025. That acceleration — 230% growth in approximately six months — is the underlying reality driving both the business adoption numbers and the capital commitment that followed.

The $100 Billion Infrastructure Bet

The new Amazon deal is structured as a 10-year, $100 billion commitment by Anthropic to AWS compute technologies, covering everything from Amazon’s Graviton CPUs to its Trainium2 and Trainium3 AI accelerators, with options on future generations of custom Amazon silicon. In exchange, Amazon is injecting a further $5 billion in new investment into Anthropic, with up to $20 billion more available over time.

The headline number is 5 gigawatts of committed capacity — a quantity that demands context. Five gigawatts is approximately the power consumption of a city the size of Los Angeles. It dwarfs any previously disclosed AI infrastructure commitment from a single model developer. Anthropic’s current deployment already runs on over one million Trainium2 chips, and the deal commits to nearly one gigawatt of Trainium2 and Trainium3 capacity by the end of 2026, with more to follow in 2027 and 2028.

“Anthropic’s commitment to run its large language models on AWS Trainium for the next decade reflects the progress we’ve made together on custom silicon,” said Amazon CEO Andy Jassy.

Project Rainier — an Anthropic-Amazon joint compute cluster already described by both companies as one of the largest in the world — is the physical manifestation of the longer-term deal. Its purpose is to give Anthropic training independence at a scale that makes pursuing the next generation of Claude models financially and logistically feasible without being held hostage to Nvidia GPU availability. Amazon’s Trainium chips, which the company has claimed offer over 30% better price-per-token than competing hardware, are central to that calculus.

Why the Amazon Relationship Is Structurally Different

Anthropic’s compute arrangements span all three major hyperscalers — AWS, Google Cloud, and Microsoft Azure — but the Amazon relationship is qualitatively distinct. The original deal, signed in 2023, was the first major outside capital Anthropic raised after its founding by Dario Amodei, Daniela Amodei, and colleagues who left OpenAI. Amazon was not just a compute vendor; it was a foundational strategic partner at a moment when Anthropic’s survival was not guaranteed.

The expansion to $100 billion represents both the fulfillment of that original bet and a significant deepening. For Amazon, the deal solves a critical problem: AWS has historically lagged Google Cloud and Azure in AI developer mindshare. Claude on Bedrock gives AWS an anchor AI product that is genuinely competitive at the frontier — not a reseller of someone else’s model, but a deep partnership with a company Amazon has meaningful ownership of.

For Anthropic, the deal provides something more important than money: certainty of compute supply. The central bottleneck for every frontier AI lab is not capital — there is more capital available than can be deployed usefully — but physical compute infrastructure. GPUs are constrained by chip fabrication capacity at TSMC and Samsung. Amazon’s custom Trainium silicon represents an alternative path that sidesteps Nvidia scarcity, and locking in 5GW of committed capacity gives Anthropic a training and inference runway that its competitors cannot easily replicate.

The Revenue Recognition Controversy

A subplot that analysts are watching closely involves how Anthropic accounts for cloud partnership revenue. According to reporting from The Information, Anthropic includes gross revenue from its cloud partnerships — including the amounts Amazon pays to host Claude on Bedrock — in its headline revenue figures, while OpenAI reports net figures that exclude such payments. Bank of America estimated that cloud partnership payments flowing through these arrangements could reach $6.4 billion annually.

The discrepancy matters because both companies are heading toward public markets, and revenue figures will be scrutinized intensely in S-1 filings. Anthropic filed confidentially for its IPO in early June 2026 at a reported $965 billion valuation — a number that, on gross revenue accounting, implies a price-to-sales multiple roughly comparable to OpenAI’s at its last private valuation. On a net revenue basis, the multiple would be substantially higher, potentially complicating Anthropic’s public market positioning.

What This Means for OpenAI

The Ramp crossover is a data point, not a verdict. OpenAI still serves far more total AI interactions, and its ChatGPT product retains commanding consumer market share — approximately 54.7% of global chatbot web traffic according to recent estimates, versus Claude’s 8.2%. The consumer market and the enterprise market are not the same competition.

But enterprise revenue is structurally stickier, tends to generate higher margins, and is the primary driver of the valuation multiples that will define the upcoming AI IPO wave. OpenAI’s own enterprise push — including its DeployCo consulting subsidiary and deepened Microsoft Copilot integration — reflects awareness that the enterprise race is the one that determines long-term economics.

The more significant implication of Thursday’s data is the speed of the reversal. Six months ago, the enterprise AI conversation was primarily about whether any company other than OpenAI could achieve meaningful penetration. Today, that question appears to have been answered.