Oracle's AI Expansion Rests on Two Unresolved Questions: Who Pays for the Chips, and Who Powers Them

Oracle is discussing investor-funded AI chip purchases with Apollo and Goldman Sachs, while power deals and OpenAI revenue questions shape the stock debate.

Oracle is exploring a financing arrangement that would keep a major purchase of AI chips off its own balance sheet, according to a Wall Street Journal report from October 7. Under the structure being discussed, an investor-funded vehicle would acquire the hardware and lease it back to Oracle. Apollo and Goldman Sachs are named as parties to the talks, which remain ongoing — no completed agreement can be inferred from them.

The distinction matters for anyone weighing the stock. Splitting ownership of the chips from their use is not simply a procurement exercise; it is a question of how Oracle finances its access to computing capacity. Tying outside investors into the chip purchase represents one possible route, but it is still under negotiation rather than an executed transaction. Talks with financing partners are not commitments, and the discussed structure does not mean Oracle has already bought the chips or signed a lease. What investors are really assessing, then, is whether the model gets implemented — not a buildout that has already happened.

Powering the Buildout Is a Separate Hurdle

While the financing question simmers, Oracle is dealing with the more mundane problem of keeping its data centers supplied. Bloomberg reported Thursday that the company is trucking compressed natural gas to sites so that work can proceed despite delays in pipeline infrastructure. Oracle is weighing the same approach for Project Jupiter in New Mexico.

A second supply arrangement covers Wisconsin. On October 2, Oracle agreed with We Energies to draw 125–250 megawatts from the Point Beach nuclear plant, subject to regulatory approval. That deal adds a concrete infrastructure component to the financing story, though the two threads sit at different stages: the chip financing is still being discussed, while the power purchase is agreed and awaiting a green light. Customers of the utility are expected to see $300 million in fuel-cost savings over 2027 to 2033 — a projection that applies to supply customers, not a guaranteed financial benefit for Oracle.

OpenAI's Revenue Picture Reshapes the Debate

The stock fell 2.5% in the prior session amid debate over OpenAI's revenue. Media reports pointed to concerns about whether OpenAI can meet its obligations to Oracle, and that is where the pivotal investor question lies: it is not only the capability of AI offerings that counts, but the economic staying power of the customers behind them.

Reuters reported Thursday that OpenAI told investors its annualized revenue was just under $50 billion at the end of September. Earlier media accounts had put the figure at roughly $70 billion. The gap shifts attention to the basis for previous expectations — though a lower number than previously reported does not by itself prove a business downturn. What is being compared are different revenue figures, not necessarily consecutive business results, and annualized revenue is not the same as revenue already booked for a year.

For Oracle, that line-drawing is more than semantics. The selling pressure was tied to doubts about OpenAI's ability to honor commitments — a worry about future payments, not confirmed news of a default. Treating the two as equivalent turns a valuation risk into an event that has already occurred. Here lies the boundary between AI enthusiasm and durable investor analysis: technological progress can support expectations, but it does not automatically answer how reliably a customer funds its promises.

Demand Signals Are Not Payment Proof

Another report underscores the same tension, with Oracle named as a cloud provider. According to media accounts, the matter involves a five-year lease for about 100,000 AI chips with a reported volume of roughly $7 billion. Oracle did not confirm the announcement. An investor should read that report as a signal of possible demand, not as secured evidence of corresponding revenue — the same distinction that applies to OpenAI. A large commitment and its economic fulfillment are two different things.

That leaves progress on the offering side intact without resolving the financing question on the customer side. The common thread running through these developments is not a blanket verdict on AI. At Oracle, the investor's focus lands on the link between technical capability, demand, and the ability to pay. Product advances remain relevant, but for framing the recent share weakness, what counts is which expectations about OpenAI hold up — and which rested only on earlier revenue reports.