OpenAI and Anthropic’s IPO race puts AI revenue quality on the clock

Anthropic and OpenAI are moving toward public markets in a race that could force investors to put hard valuation rules around the AI-lab business model. Reuters reported that Anthropic announced on June 1, 2026 that it had made a confidential filing with U.S. regulators, and that OpenAI followed about a week later. The stakes are larger than which company lists first: Reuters reported that the companies view a first listing as a way to frame how investors value them, making the first prospectus a likely reference point for AI revenue, costs, partner dependence, and control.

The timing compresses a private-market story into a public-market test. Reuters reported, citing two people familiar with the matter, that OpenAI told some investors it was targeting an IPO as early as September 2026. A separate Reuters account said investors are scrutinizing whether the AI sector’s rise can be sustained, and reported, citing a source familiar with the matter, that OpenAI told investors during its most recent fundraising round that it did not expect to be profitable until 2030. Those claims are not public company guidance in the supplied evidence, but they make profitability timing part of the listing question.

The clearest issue is not whether AI demand exists; it is what kind of revenue public investors are being asked to buy. MediaPost reported that Anthropic and OpenAI share material revenue amounts with business partners, and said Anthropic shares revenue with Amazon and Google, which resell its models. That supports a revenue-quality question, but not a settled accounting conclusion: the research pack includes no S-1, audited financial statements, company disclosures, or authoritative accounting analysis showing how the companies recognize partner-distributed revenue.

OpenAI’s partner economics are also unresolved in the supplied evidence. MediaPost reported that OpenAI renegotiated agreements to allow partnerships with Amazon and Google while continuing a Microsoft revenue share through 2030. A second MediaPost item reported that the revised Microsoft relationship changed revenue-share economics tied to Microsoft cloud sales and that OpenAI can sell models on AWS and Google Cloud without Microsoft’s permission. The two accounts point in the same broad direction, toward less Microsoft exclusivity, but they leave the direction, scope, duration, and payment mechanics of the revenue-sharing arrangements unclear.

That is why the first prospectus matters. Private AI financings can lean on growth, strategic backing, and future capacity; a public filing would have to describe revenue recognition policies, gross margins, related-party exposure, customer concentration, cloud commitments, and risk factors in a form investors can compare. CNBC’s Facebook post framed future OpenAI and Anthropic prospectuses as requiring investors to understand a new economy, including token economics. That is landscape framing rather than proof of a particular accounting treatment, but it captures the likely investor-learning curve.

For now, the strongest supported conclusion is narrower than a verdict on AI margins. Reuters supports the claim that Anthropic and OpenAI are racing toward IPOs and that first-listing status is seen as a way to shape valuation norms. Reuters also supports, with anonymous sourcing, that OpenAI has discussed a September 2026 target with some investors and has told investors it does not expect profitability until 2030. MediaPost supports only an attributed concern that partner revenue-sharing and cloud resale relationships may complicate headline revenue.

The next evidence threshold is public documentation. If either company releases an S-1 or equivalent prospectus, the story moves from rivalry and investor messaging to accounting detail: who owns the customer relationship, whether partner-sold model revenue is presented gross or net, how much compute and cloud spending sits behind each dollar of revenue, and how much control cloud partners retain. Until then, the IPO race is best read as an early test of whether public markets will value AI labs like software platforms, infrastructure-intensive service providers, or something less cleanly comparable.