As Anthropic Closes In on Profitability, OpenAI Files for IPO

Anthropic told investors on 20 May 2026 that it expects to report its first profitable quarter, with second-quarter revenue on track to exceed $10.9 billion — more than double the prior period's take, according to reporting by TechCrunch. The milestone, if confirmed in full, would mark a decisive shift for a company that has burned through billions in compute costs while building its Claude family of models.

The timing matters. Polymarket, the decentralised prediction market, opened a market that same day asking which of the two AI heavyweights — Anthropic or OpenAI — would hold the higher valuation at the close of June 2026. Within hours, the market was pricing an 83 percent probability on Anthropic prevailing. A separate market put the odds of an OpenAI AGI announcement before 2027 at just 12 percent. Both signals reflect something the financial press has been slower to absorb: the AI race is no longer measured in benchmark scores alone.

What changes when an AI lab turns profitable

The profit inflection point carries more weight than a balance-sheet footnote. Anthropic's revenue trajectory, if the $10.9 billion Q2 figure holds, places it in the upper tier of enterprise software companies by annualised sales — a category in which pure-play AI firms have been conspicuous by their absence. The path from research laboratory to self-sustaining business has historically demanded either a platform transition (the cloud infrastructure model) or a distribution moat (the enterprise software lock-in). Anthropic appears to be betting on the former through its partnerships with Amazon and Google, and on the latter through Constitutional AI, a governance framework pitched to regulated industries as a compliance advantage rather than a liability.

The question is whether that pitch survives contact with OpenAI's commercial machinery. OpenAI has operated at a far larger revenue scale for longer, but has not publicly disclosed profitability. Sources cited by Polymarket trackers and financial wire services on 20 May 2026 indicated that OpenAI is preparing to file for an initial public offering in the coming days. An IPO would impose disclosure obligations that a private lab can avoid — including detailed breakdowns of revenue, cost of compute, and the structural subsidy embedded in the non-profit's commercial arm.

The math problem that wasn't

Before the commercial narrative can be told cleanly, the research narrative requires its own accounting. On 20 May 2026, OpenAI announced that one of its internal reasoning models had resolved a geometry conjecture first posed in 1946 — a result that, if valid, would represent one of the more striking examples of AI-assisted mathematical discovery. The announcement carried particular salience because OpenAI had previously published a mathematical claim later retracted after independent researchers identified critical errors. This time, OpenAI pointed to mathematicians who had previously challenged its work as the validators of the new result.

The episode illustrates an asymmetry in how AI research claims circulate. When an AI lab announces a benchmark-beating performance or a formal proof, the verification community — academic mathematicians, independent researchers — operates on a slower cadence than the announcement machinery. For OpenAI, the reputational cost of a second disputed claim would be substantial. For the field broadly, it raises a structural question about the epistemic infrastructure available to distinguish genuine mathematical discovery from pattern-matching that resembles it.

The valuation signal and what the market is actually pricing

The 83 percent Polymarket probability is not a neutral forecast. It reflects a market populated by traders who have priced in a version of the story in which Anthropic's revenue acceleration, combined with its structural independence from the non-profit governance complications that continue to shadow OpenAI, makes it the more credible long-term vehicle for institutional capital. OpenAI's IPO filing, if confirmed in the coming days, would represent a bet on the public market's appetite for a company whose valuation has historically been held aloft by capability claims rather than audited earnings.

That appetite is not guaranteed. The technology sector's public history contains numerous examples of companies that scaled rapidly in private markets and found the public equivalent a harder crowd. The difference, and it is not a minor one, is that the AI sector has demonstrated genuine revenue growth — not merely the promise of it. If Anthropic's $10.9 billion Q2 materialises and OpenAI files its S-1 within the projected window, the next six weeks will offer a natural experiment in how capital allocates between a lab transitioning to profitability and one transitioning to public markets.

What remains unclear

The sources do not specify the exact date Anthropic expects to close the Q2 reporting period, nor do they confirm the precise structure of OpenAI's IPO filing — whether it intends a direct listing, a traditional underwritten offering, or some hybrid structure. The valuation figure being priced by the Polymarket market reflects a composite estimate that aggregates anonymous trader sentiment rather than audited data. And the mathematicians who have reportedly validated the 1946 conjecture result have not published their assessment in a peer-reviewed venue as of this writing.

Anthropic declined to comment beyond its investor communications. OpenAI did not respond to requests for comment on the IPO timing or the mathematical research claim.

The desk chose to lead with Anthropic's financial milestone rather than OpenAI's IPO filing on the grounds that a first profitable quarter is a categorically different kind of news than a preparatory regulatory step. The Polymarket signals, while not independently verifiable, offered a useful proxy for how informed traders are positioning ahead of what may prove to be a pivotal quarter for the industry's commercial maturation.