OpenAI completed an approximately $7 billion employee share buyback on August 10, 2026, with its valuation locked at $852 billion — the same figure set by its record $122 billion funding round that closed on March 31. Bloomberg first reported the transaction. The details that matter most aren’t the numbers. They’re who paid, and what didn’t happen.
OpenAI paid for the buyback itself, out of its own cash. No Thrive Capital, no SoftBank, no outside investors. And the valuation didn’t move — no premium, no discount, a flat print across the company’s largest liquidity event to date. For a company this closely watched, a flat valuation is a statement: the private benchmark holds, and there’s no rush to reprice it in public.
The deal in numbers
This is the third large liquidity event OpenAI has run in roughly two years, and each one marks a step-change in how the market values the company:
- 2024: $1.5 billion employee tender, valuation not disclosed.
- Before the 2026 rounds: $6.6 billion secondary sale at a $500 billion valuation.
- March 2026: $122 billion primary round at $852 billion post-money — the largest private funding round in history, co-led by SoftBank alongside Amazon and Nvidia.
- August 2026: $7 billion buyback at $852 billion, self-funded.
The transaction covered shares held by current and former employees. Large private AI companies routinely let insiders cash out portions of their holdings; OpenAI had previously done it through external investors. Doing it with company cash changes the capital structure of the event: no new money, no new voices at the table, no dilution.
What the flat print signals
A buyback at an unchanged valuation, paid for internally, reads as confidence mixed with patience. Confidence, because writing a $7 billion check to your own people from your own balance sheet is not something a cash-strapped company does. Patience, because the surest way to push a valuation up would have been to bring in new outside capital at a higher number — and OpenAI chose not to.
The unchanged valuation also reflects a market that has stopped automatically repricing OpenAI upward. Competition in generative AI has intensified — DeepSeek just fired the latest shot in the open-weights price war — multiple labs are shipping frontier models on faster cycles, and investors have grown more cautious about any single company’s long-term dominance. The $852 billion mark holding flat across the tender suggests the company’s revenue and growth expectations — Bloomberg reported in August that annualized revenue had surpassed $40 billion, roughly double the run-rate at the close of 2025 — support the number without demanding a new one.
The IPO question
The tender runs in parallel with OpenAI’s path to public markets. The company confidentially filed its S-1 registration statement with the SEC on June 8, with Goldman Sachs and Morgan Stanley as lead underwriters — the standard first step toward a possible listing. The buyback doesn’t change that filing. What the flat valuation signals is that no debut is baked into near-term planning.
The timeline has been slipping in plain sight. The original target for the listing was September 2026. But on August 19, CFO Sarah Friar told employees at an all-hands that OpenAI “will be a public company in 2027,” or sooner if the business kept inflecting. Then CEO Sam Altman confirmed the delay himself: in a September interview, asked whether OpenAI would list in 2026, he said plainly that right now would be an ill-advised moment to go public.
A confidential S-1 starts the SEC review clock without fixing a date, and it’s common for companies in that quiet window to run tenders and tidy up their ownership before terms are set. OpenAI’s buyback fits that pattern exactly. The message to employees: you can get liquidity now, at the number we all agreed on in March. The message to the market: we’ll come when we’re ready, and the number starts with a T. The same discipline is visible in the product org, which just shut down its ChatGPT Atlas browser to focus on fewer, bigger bets.
Inside the $122 billion round
To understand the buyback, rewind to March 31 — the $122 billion round that set the $852 billion benchmark. SoftBank co-led alongside Amazon and Nvidia, with Amazon committing $50 billion (of which $35 billion is contingent on OpenAI going public or hitting a specified AGI milestone) and Nvidia and SoftBank each putting in $30 billion. Notably, OpenAI opened approximately $3 billion of the round to individual investors through bank channels — a structural first that widened the shareholder base beyond institutions.
The round’s scale reflected a simple reality: frontier AI is the most capital-intensive business ever built by a startup. Training runs, inference fleets and the talent to build them consume cash at a rate that makes traditional venture math look quaint. The $122 billion wasn’t a war chest in the usual sense; it was the entry fee for the next phase of the scaling race.
The $600 billion compute bill
That capital has a destination. The new funding extends a compute commitment approaching $600 billion touching Microsoft Azure, Oracle, Amazon Web Services, CoreWeave and Google Cloud — the largest infrastructure procurement program in private-market history. The chip side of that buildout just sent AMD past a $1 trillion market cap on the back of OpenAI and Oracle orders. Every dollar of it is a bet that demand for AI compute keeps growing faster than supply, and that owning the capacity is the moat.
This is the context in which the $7 billion buyback’s self-funding matters. A company writing nine-figure checks to cloud providers while simultaneously writing a $7 billion check to its own employees from operating cash is signaling something specific: the machine generates enough cash to fund both growth and liquidity. That’s a meaningful statement for a company that has never been profitable in the conventional sense.
The valuation debate
Not everyone is convinced the $852 billion number is solid ground. A Financial Times investigation in April surfaced what some institutional investors call a “valuation gap”: Bridgewater partner Greg Jensen reportedly told clients the implied 35-times forward revenue multiple is “priced for a monopoly outcome that does not yet exist.” The flat tender print doesn’t resolve that debate — it just shows OpenAI isn’t interested in having it right now. With annualized revenue past $40 billion and the IPO officially a 2027 story, the company has bought itself time. Time, at this valuation, is the most expensive thing it owns.
Featured image: OpenAI via Wikimedia Commons (Public domain (PD-textlogo))
