AI's Profitable Moment: Anthropic's Revenue Surge Meets OpenAI's IPO Filing

On 20 May 2026, OpenAI filed preliminary documentation for a public offering, according to reports cited across prediction markets and financial commentary. That same day, Anthropic disclosed to investors that it expects to more than double its revenue to approximately $10.9 billion in the second quarter — and that this will constitute its first profitable quarter, according to reporting by TechCrunch. The two disclosures, landing within hours of each other, crystallise a moment the AI industry's backers have been building toward: the transition from speculative research bets to companies expected to justify their valuations through audited earnings.

The collision of Anthropic's revenue milestone with OpenAI's IPO paperwork creates an unavoidable comparison. Both companies have raised billions at valuations that presuppose revenues far beyond what any software business has historically delivered at their stage. Anthropic's Q2 projection suggests those assumptions may be within reach — or at least within sight. OpenAI's readiness to enter public markets implies its own financial trajectory has satisfied the gatekeepers who advise on such transitions. What neither disclosure resolves is which model — heavily regulated safety-first research, or broadly deployed consumer and enterprise products — will command the higher multiple once the market has seen both sets of books.

The Math That Landed Wrong, Then Right

The timing of OpenAI's mathematical announcement also drew scrutiny. The company claimed its internal reasoning model had successfully disproved a geometry conjecture that had remained unsolved since 1946. This was not the first time OpenAI has made a headline-grabbing capability claim. An earlier, highly publicised mathematical assertion had been exposed as erroneous by external reviewers — an embarrassing episode that complicated the company's credibility on technical claims. In this instance, however, mathematicians who had previously identified OpenAI's errors were among those consulted and reportedly confirmed the revised result. The shift matters: OpenAI's research announcements now undergo a form of external pressure-testing before they are treated as settled. Whether that process is genuinely independent or merely performative remains contested in research communities, but the company's willingness to subject itself to review — and the reviewers' willingness to affirm the outcome — is a notable data point on how AI capability claims are being audited at the frontier.

The Valuation Wager

Prediction markets reflect the genuine uncertainty in where this race ends. Polymarket hosts an open contract asking which company — Anthropic or OpenAI — will carry the higher valuation at the close of June 2026. Separately, markets assign approximately a twelve percent probability to OpenAI announcing it has achieved artificial general intelligence before the end of the calendar year. That number is low enough to signal widespread scepticism, yet high enough to suggest a non-trivial body of participants considers the claim plausible. The coexistence of an IPO filing with a twelve-percent AGI probability reveals how the industry is pricing uncertainty: the commercial infrastructure is being built as if one future is certain, while the technical definition of that future remains contested.

The Structural Shift

What is happening to both companies is also a structural transformation of the AI research model itself. Anthropic and OpenAI were founded as organisations that would resist the commercial incentives to deploy capabilities before they were understood. Both have moved, with varying degrees of explicitness, toward revenue targets, enterprise contracts, and — in OpenAI's case — a public shareholders base. The result is a genuine tension: the research questions they were constituted to pursue do not resolve on a quarterly earnings calendar. Investors underwriting an IPO or a revenue projection need predictability. The history of AI development suggests that breakthroughs arrive in irregular bursts, not steady increments. Both Anthropic and OpenAI are navigating the gap between what they were built to do and what the capital markets now require them to demonstrate.

Stakes

The outcomes here are not symmetrical. If Anthropic sustains profitability while OpenAI stumbles in public markets, the safety-first research model gets a powerful vindication — and a harder-to-ignore seat at the table when regulators set rules for frontier AI. If OpenAI's IPO succeeds and its revenue scales, the commercial-deploy-and-monetise model becomes the template for every startup that follows. For enterprise buyers, the difference may be contractual: OpenAI has been willing to move fast on product; Anthropic has built its brand on refusing to do so until certain conditions are met. Neither approach is obviously wrong. Both are being stress-tested simultaneously.

Market-implied odds on OpenAI achieving AGI this year sit at roughly twelve percent — a figure that reflects genuine uncertainty rather than dismissal. The coming months, as both companies cross their respective thresholds, will clarify which model the capital markets actually reward.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://x.com/unusual_whales/status/1921572345678282952