Anthropic 965B Valuation Surpasses OpenAI in 65B Round

The number that should stop you is not the $65 billion Anthropic just raised. It is the $47 billion revenue run rate the company disclosed in the same breath. A research lab that most enterprise buyers had not heard of three years ago is now generating annualized revenue that rivals the entire software business of SAP, and it did so largely on the back of a single product line: code.

What Actually Happened

On Thursday, May 28, Anthropic confirmed it had closed a financing round of $65 billion, valuing the company at $965 billion including the new capital. The round was led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, a roster that reads like a who's who of growth investors who spent the prior decade backing the SaaS winners that AI now threatens to absorb. The valuation pushes Anthropic past OpenAI for the first time, an inversion almost no one outside the company's cap table predicted even six months ago.

Anthropic paired the raise with two product disclosures. The first was Claude Opus 4.8, its newest frontier model, shipped the same day. The second was the existence of Claude Mythos Preview, a model with advanced offensive and defensive cybersecurity capabilities that the company is deliberately withholding from general release, offering it only to a vetted set of organizations. Alongside the round, Anthropic reported a $47 billion revenue run rate, a figure driven overwhelmingly by Claude Code, its agentic software engineering product, and by enterprise API consumption.

Why This Matters More Than People Think

For two years the AI narrative had a fixed center of gravity: OpenAI was the franchise, everyone else was the challenger. That framing organized everything from talent flows to media coverage to enterprise procurement defaults. A $965 billion valuation that explicitly clears OpenAI's last marked round breaks the frame. It tells the market that the leader can change, and that the thing separating first from second is no longer brand recognition but revenue durability and enterprise trust.

The competitive consequence is immediate. Enterprise buyers who standardized on a single vendor now have cover to run a real bake-off, because the "obvious safe choice" argument just lost its anchor. Cloud providers that hedged across multiple labs look prescient. And the capital markets, which had been pricing AI leadership as a winner-take-most outcome, now have to price a genuine duopoly at the frontier, with a long tail of open-weight challengers underneath.

The Competitive Landscape

OpenAI is not standing still. It recently raised $122 billion of its own and is pushing aggressively into consumer surfaces, advertising, and a rumored IPO. But the two companies are increasingly fighting different wars. OpenAI's revenue base skews toward ChatGPT consumer subscriptions and a sprawling product surface, while Anthropic's revenue is concentrated in developer and enterprise spend where switching costs, compliance posture, and reliability matter more than viral reach. That concentration is both Anthropic's strength and, as critics argue below, its exposure.

Google sits in an unusual position: it is simultaneously Anthropic's largest infrastructure partner, having committed multiple gigawatts of next-generation TPU capacity, and a direct competitor through Gemini. Amazon, too, has tied itself to Anthropic with a chip and cloud arrangement reportedly exceeding $100 billion over ten years. The result is a strange web in which the hyperscalers are underwriting the very lab that competes with their own foundation models, a bet that only pays off if Anthropic's enterprise revenue keeps compounding. Meta and xAI, meanwhile, are attacking from the open-weight and price-war flank, undercutting both leaders on licensing cost.

Hidden Insight: Anthropic Sold Trust, Not Tokens

The conventional reading is that Anthropic won because Claude is a better coding model. That is partially true, but it misses the deeper move. Anthropic built its entire commercial identity around safety, interpretability, and predictability, and for years that looked like a tax on growth, a reason it would always trail flashier rivals. The $965 billion valuation reveals that the safety posture was not a tax. It was the product. Enterprise buyers in regulated industries, banks, insurers, healthcare systems, defense contractors, do not buy the cleverest model. They buy the model their compliance team will sign off on. Anthropic spent years making itself that model.

Claude Mythos sharpens the point. By building a model with elite cybersecurity capability and then refusing to ship it broadly, Anthropic is performing restraint as a market signal. To a CISO deciding which lab to trust with sensitive infrastructure, a vendor that visibly withholds its most dangerous capability is more credible than one racing every capability to market. Restraint, counterintuitively, has become a sales asset. Over the next 12 to 24 months, expect this to harden into a moat: the labs that can credibly say "we held something back" will win the highest-margin, highest-trust contracts, while the labs optimizing purely for benchmark supremacy compete on price.

The uncomfortable truth this challenges is the assumption, dominant since 2023, that raw capability is the only variable that matters. It never was. The model that wins the enterprise is the one a general counsel will defend in a deposition, and that is a fundamentally different selection criterion than topping a leaderboard.

The Bear Case Worth Taking Seriously

The bear case is straightforward and should not be waved away. A $965 billion valuation against a $47 billion run rate is roughly 20 times forward revenue for a business whose revenue is dangerously concentrated in code generation, the single category where commoditization is moving fastest. However impressive Claude Code's growth, open-weight models from Chinese labs and Meta are closing the quality gap while pricing at a fraction of frontier cost. If a "good enough" coding model becomes free, Anthropic's highest-growth revenue line faces brutal margin compression. Skeptics also point out that much of Anthropic's revenue flows through cloud partners who are themselves competitors, giving those partners both visibility into Anthropic's economics and the incentive to substitute their own models over time. The risk the market may be underpricing is correlation: nearly every dollar of Anthropic's valuation, its compute supply, and its distribution runs through two or three hyperscalers whose interests will not always align with Anthropic's.

What to Watch Next

Watch three indicators over the next 90 days. First, the gross margin Anthropic discloses or leaks on Claude Code: if it holds above 60 percent against open-weight pressure, the moat thesis survives; if it compresses toward 40, the bear case is winning. Second, enterprise logo concentration: a follow-the-money look at whether the $47 billion run rate is broad-based or propped up by a handful of mega-contracts with the hyperscalers and a few banks. Third, OpenAI's response, which will likely come as either an aggressive enterprise price cut or an accelerated IPO timeline meant to reframe the leadership narrative before it calcifies.

Over 180 days, the decisive question is whether the Mythos restraint strategy converts into named, marquee government and defense contracts. If Anthropic announces a major national-security deployment built on its "we withheld it" credibility, the trust-as-product thesis goes from interesting to proven. If Mythos stays a press release with no contracts attached, it was theater. If you are an investor, track the cloud partners' own model roadmaps as a leading indicator: the day Google or Amazon positions a first-party model directly against Claude for enterprise coding is the day the duopoly's friendliest alliance turns adversarial.

Anthropic did not win the AI race by building the smartest model. It won by becoming the only model a Fortune 500 general counsel is willing to defend in court.

---