The Money Turn — OpenAI Readies a Record IPO as Anthropic Passes It on Revenue
OpenAI is preparing a confidential IPO filing targeting a debut as soon as September at a roughly 730 billion dollar valuation. In the same week, Fortune reported Anthropic has overtaken it on revenue, and the Federal Reserve stood up its first AI task force. The AI story just became a capital-markets story.
For three years the AI narrative has been a product narrative — bigger models, new capabilities, the next benchmark. This week it turned into a capital-markets narrative, and the two are not the same story.
The proximate event is OpenAI preparing a confidential IPO filing with Goldman Sachs and Morgan Stanley, targeting a public debut as soon as September at a private-market valuation around 730 billion dollars. If it prices anywhere near that number it becomes the largest technology IPO in history by a wide margin, and one of the defining market events of the decade. A confidential filing is still a preparation, not a bell rung on an exchange floor — but the mechanics of underwriter selection and timing signal that the company now sees public markets, not another private round, as its next funding surface.
The subplot is the more interesting number. Fortune reported this week that Anthropic has overtaken OpenAI on revenue, running at roughly 47 billion dollars annualized against OpenAI's projected 25 to 33 billion for 2026. For a company that has trailed OpenAI on public mindshare for its entire existence, a revenue lead going into a rival's IPO is an awkward fact for the roadshow to explain.
Annualized revenue, reported vs projected (USD billions)
| company | annualizedRevenueB |
|---|---|
| Anthropic (reported run-rate) | 47 |
| OpenAI (2026 projection, high) | 33 |
| OpenAI (2026 projection, low) | 25 |
The gap matters because the two companies are converging on the same buyers from opposite directions. OpenAI built a consumer brand and is monetizing enterprise underneath it; Anthropic went enterprise-first and is now the default for a large share of production agent and coding workloads. A revenue lead suggests the enterprise-first bet is compounding faster than the consumer-brand premium — and it lands just as OpenAI asks public investors to underwrite that premium.
The macro layer arrives
The third development is the one that will still matter in five years. The Federal Reserve stood up its first formal body dedicated to studying AI's impact on jobs, productivity, and monetary policy, co-led per this week's reports by a16z co-founder Marc Andreessen. It is the first time the central bank has built a standing structure around AI-driven economic effects, and the appointment of a prominent AI investor to co-lead it is already drawing conflict-of-interest scrutiny.
Set the governance question aside and the signal is unambiguous: AI has crossed from a sector story into a macro variable the Fed feels it has to model. That is the same threshold that turned the internet from a NASDAQ subplot into a productivity-statistics debate in the late nineties. When the central bank builds a task force, the technology has stopped being something markets price in isolation.
Around all of this, the plumbing kept setting records of its own — SK Hynix's Nasdaq debut earlier in the week raised roughly 28 to 29 billion dollars, the largest American depositary receipt listing in history, a memory-maker riding the same demand curve that our analysis of the physical layer of AI traced from models down to megawatts.
Why the sequencing is the story
Read the three events as a sequence and a pattern emerges. A revenue lead changing hands, a record IPO being staged, and a central-bank task force forming are not three unrelated headlines — they are the same transition seen from three altitudes. The capital that has been flowing into AI privately is now looking for public exits, public benchmarks, and public accountability, all at once.
For enterprise buyers the practical takeaway is about leverage. When the vendor you depend on is heading for public markets, its pricing and roadmap start answering to quarterly investors, not just to your renewal. That is the same pressure our reporting on the quiet repricing of AI token consumption has tracked from the buyer side, and it is why our standing prediction on enterprise AI consolidation by 2027 looks more, not less, likely after a week like this one. The talent fights already in the open — captured in our coverage of Apple's trade-secret suit against OpenAI — are the human-capital version of the same scramble now playing out in the capital-markets version.
The product race is not over. But this week it stopped being the only race that sets the price.
Sources