OpenAI’s revenue debate raises a bigger question for the companies building its computing infrastructure: how much future business can their spending support?
A difference between gross and net revenue can make two measures of the same business look far apart. It does not, by itself, show that customers disappeared or that sales missed a forecast. For Oracle, CoreWeave and the chipmakers tied to AI expansion, the more consequential issue is whether long-term commitments become profitable business.
Revenue comparisons need the same yardstick
For sales routed through a partner, a gross revenue measure can include the partner’s share, while a net measure excludes it. Comparing OpenAI with Anthropic therefore requires consistent treatment of those sales. Otherwise, the headline figures can exaggerate the distance between the businesses.
Annualized revenue adds another distinction. It projects a shorter period’s sales over a year, rather than measuring sales already booked across a complete financial year. It also says nothing on its own about profit or cash available to spend.
That makes an apparent revenue gap a poor shortcut for judging whether OpenAI can afford its computing commitments. The useful comparison needs matching accounting definitions, clear time periods and a separate look at costs.
Oracle and CoreWeave face the delivery test
Oracle’s cloud expansion illustrates the financial stakes. In its September 10, 2026 earnings announcement, the company reported $664 billion in remaining performance obligations, a measure of contracted future revenue. It also reported roughly $5 billion in negative free cash flow for its fiscal first quarter as it invested in cloud infrastructure.
Those figures describe Oracle’s overall business. They do not establish how much of its backlog depends on OpenAI. But they show why a large order book and pressure on cash flow can coexist: delivering future services requires spending along the way.
CoreWeave offers a more specific view of its OpenAI relationship. In September 2025, the company put the total potential value of its OpenAI agreements at up to approximately $22.4 billion. That ceiling describes contract value, not money already collected.
CoreWeave subsequently reported $5.13 billion in revenue for 2025 alongside a net loss. Its challenge is to turn contracted demand into profitable operations. A large customer commitment can support expansion without settling whether the infrastructure will earn an adequate return.
The exposure extends to chipmakers
OpenAI and Broadcom announced plans in October 2025 to co-develop systems using custom AI accelerators. AMD and OpenAI also announced a multiyear GPU deployment agreement that month. These arrangements give both suppliers a stake in OpenAI’s infrastructure ambitions.
Nvidia could also feel a slowdown if cloud providers reduced their hardware purchases. That is a potential consequence of weaker infrastructure spending, rather than evidence that orders have already been cut.
The distinction matters across the supply chain. Demand for computing capacity can support chip sales while leaving the companies financing and operating that capacity with a separate profitability problem.
The key question is whether the spending pays off
The risk is that infrastructure spending gets ahead of the revenue needed to sustain it. If customers expand more slowly than expected, providers could face expensive capacity that takes longer to earn its keep.
That possibility deserves scrutiny without treating mismatched revenue measures as proof of a collapse. OpenAI’s accounting presentation and its ability to fund future computing needs are related questions, but they require different evidence.
For investors assessing these deals, the meaningful signals are revenue earned on a consistent basis, cash generation and the cost of fulfilling contracts. Big commitments establish the scale of the ambition. The returns still depend on execution.