Investors have pitched OpenAI a $1.2 trillion valuation. OpenAI wants $1.5 trillion. Neither number comes with a public IPO date.

Investors have been sounding out OpenAI on a new private funding round that would value the company at roughly $1.2 trillion -- OpenAI itself is said to want $1.5 trillion -- according to the Financial Times. Either number would be a sharp jump from the $852 billion post-money valuation OpenAI priced in March, when it closed a $122 billion round led by SoftBank, Amazon and Nvidia. The talks are described as early and investor-initiated, not something OpenAI went looking for, and nothing is final.

The March round is worth remembering for how it was actually built, because it shapes what a bigger one would need to look like. Alongside SoftBank, Amazon and Nvidia, the roster included Andreessen Horowitz, D.E. Shaw Ventures, MGX, TPG and T. Rowe Price Associates -- and for the first time, roughly $3 billion of the total came from individual investors through bank channels and ARK Invest-managed ETFs, rather than staying confined to OpenAI's usual circle of institutional backers. OpenAI also expanded an undrawn credit facility to roughly $4.7 billion around the same time, backed by several major global banks -- a second, quieter source of capital sitting alongside whatever a new round eventually raises.

The case for a bigger number is real. OpenAI's annualized revenue topped $40 billion last month, roughly double where it ended 2025, driven in part by July's GPT-5.6 release -- up from $5.7 billion in the first quarter of 2026 to $6.7 billion in the second, according to figures reported by Yahoo Finance. The company now counts more than a billion weekly active users and over 200 million business customers, and it ran a roughly $7 billion secondary sale in August that let employees cash out shares. None of that is in dispute. Spending has scaled just as fast: OpenAI spent $34 billion on AI model training alone last year, a number that only grows as each new model generation gets more expensive to build.

What's less settled is what that revenue is actually worth, and the clearest way to see the gap is next to the AI lab OpenAI would still be valued above. Anthropic's own annualized revenue reached $65 billion by the end of July -- 62% more than OpenAI's, off a base that was itself smaller than OpenAI's a year ago. The climb was fast even by this industry's standards: Anthropic ended 2025 at roughly $9 billion, reported $14 billion in February, $30 billion by April, $47 billion in May, and $65 billion by the end of July -- adding $18 billion in just the two months before that last mark. Anthropic's most recent reported private valuation, $965 billion in late May, sits close to OpenAI's current one; its own IPO target, as soon as October, is $2 trillion or more.

PitchBook's own scoring of the two companies lands in the same direction. Its AI Business Quality index -- built from margins, revenue concentration and customer mix, not growth alone -- rates OpenAI at 4.8 out of 10, the lowest of the major labs it scored; Anthropic comes in at 7.4, Databricks (a separate enterprise-software company, not a foundation-model lab) at 8.7, the highest in the framework. The gap traces to specifics rather than vibes: PitchBook puts roughly 85% of OpenAI's 800-900 million weekly users on free tiers that generate no direct revenue, flags OpenAI's own revenue as more concentrated in consumer subscriptions and developer API fees than Anthropic's enterprise-heavy mix, and doesn't project OpenAI reaching positive free cash flow until 2029 -- three years past even the earlier end of its own possible IPO window.

$1.2 trillion on $40 billion of revenue is a number a reader can check against precedent, which is exactly what the slider above is for -- move the valuation toward OpenAI's own $1.5 trillion ask, or the revenue toward where the company hopes it lands by year-end, and the multiple moves with it. Nothing about the arithmetic is secret. What's genuinely unresolved is which side of it -- the company's growth, or the market's appetite for paying up front for it -- is the one more likely to be wrong.

Both companies have now filed confidentially for an IPO, and both keep pushing the actual date out. Altman told Fortune this month that an OpenAI listing in 2026 would be ill-timed given open safety questions, framing it plainly: "no gamble with humanity is OK." Anthropic's own reported October target, by contrast, hasn't moved. PitchBook itself calls OpenAI a company priced on promise rather than metrics. A trillion-dollar-plus valuation doesn't resolve that. It just raises the price of believing the promise.