OpenAI plans $856bn of compute against $840bn of revenue through 2030

Compute runs past revenue by $16bn in the plan ^[13], and with the forecast burn at $278bn, everything else in the business accounts for roughly $262bn of cash out: staff, research, the cost of serving the models people actually use ^[1][16].

The money already raised covers the first stretch and no more. On the presentation's own pace, March's $122bn is gone in 2028 ^[4], which leaves about $156bn of the five-year burn unfunded ^[14]. Sold at the $1.2tn valuation investors have approached with, $156bn is about 13% of the company ^[5][15], and that price already sits 41% above the $852bn March mark ^[18]. A source close to the company told the FT that OpenAI is campaigning for an even higher valuation ^[5].

All of it depends on the revenue ramp. Going from $36bn to $350bn is 9.7 times in four years, about 76% compounded annually, four years running ^[2][17]. The cumulative figure is the harder test: strip out this year and 2030, and 2027, 2028 and 2029 have to produce $454bn between them ^[19].

Bloomberg reported that the numbers came from a private presentation in July related to a computing deal, citing a person with knowledge of the deal who requested anonymity and declined to discuss financial figures ^[11]. A compute counterparty was shown the burn during a negotiation. The FT did not name it. The FT also reported that OpenAI's ability to address its funding needs is critical to a series of financial arrangements and hardware deals it has made, and that major tech groups such as Nvidia rely heavily on contracts with OpenAI for their revenues ^[9].

OpenAI filed confidentially for an IPO in June, and Altman said on Saturday that the company would not go public in 2026 because of concerns about AI safety ^[6]; earlier this month he called a listing "ill-advised" ^[7]. PYMNTS CEO Karen Webster wrote last week: "The gap between what these companies say and what they do is the most useful data point in this whole debate." ^[12] PYMNTS said it contacted OpenAI for comment and had not received a reply ^[10].

If the round closes at $1.2tn or above, the $156bn is dilution and the 2030 figures stay untested for another two years. A slower revenue path pulls the problem into 2028 with a bigger number attached, and the raise has to exceed $156bn. Should the raise not clear at any price, the compute commitments get rewritten and the counterparties holding them find out what they were underwriting. In my view 2028 is the date that governs this, because the company's own presentation puts the March cash out then ^[4] and the gap after it is $156bn ^[14]. The thesis fails if 2027 revenue tracks the ramp: three years totalling $454bn ^[19] would make the $156bn a matter of price and dilution.