SAN FRANCISCO, February 3, 2026, 12:08 PST
- OpenAI has been weighing alternatives to some Nvidia chips for certain AI “inference” work, sources said.
- Nvidia’s CEO rejected talk of a rift and said investment discussions with OpenAI remain on track.
- Oracle is tapping debt and equity markets to fund an AI-driven cloud buildout tied to OpenAI and others.
OpenAI, the maker of ChatGPT, is unsatisfied with some of Nvidia’s latest AI chips and has looked for alternatives since last year, eight sources familiar with the matter said, potentially complicating investment talks between two of the most visible names in the AI boom. The concern centers on “inference” — when a model answers a user — as opposed to training, the heavy-lift phase that has powered Nvidia’s rise in graphics processing units (GPUs). (Reuters)
That matters now because Wall Street is watching whether the AI buildout can keep expanding without spooking investors who already see balance sheets, not software demos, as the constraint. Nvidia shares were down about 5% in midday trading, while Oracle slid nearly 5%.
Oracle, which has tied much of its cloud push to AI capacity, said it expects to raise $45 billion to $50 billion in 2026 to fund expansion of its Oracle Cloud Infrastructure business to meet “contracted demand” from customers including OpenAI, Advanced Micro Devices, Meta, Nvidia, TikTok and xAI. It said roughly half would come from equity and equity-linked financing — including mandatory convertible preferred securities (stock that converts into common shares later) and an at-the-market program (selling shares into the market over time) of up to $20 billion — with the rest from a single bond issuance. (Oracle Investor Relations)
Against that backdrop, Jensen Huang told CNBC’s Jim Cramer that Nvidia’s plan to invest in OpenAI remains “on track,” adding: “There’s no controversy at all.” (StockAnalysis)
Investors have been parsing that message since a report by The Wall Street Journal said the proposed investment had stalled, dragging on Nvidia shares. Barron’s said Nvidia stock fell 2.9% on Monday, closing at $185.61, as questions swirled around the OpenAI talks. (Barron’s)
Oracle has its own nerves to settle. Investors have scrutinized the company’s AI-driven infrastructure spending as its debt rises and its fortunes become more closely tied to OpenAI, which is not profitable and has not detailed how it will finance its infrastructure plans, a Reuters report said. Oracle was sued in January by bondholders who said it concealed the need to sell significant additional debt, and the cost of insuring Oracle debt against default jumped to a five-year high in December. (Reuters)
On Monday, analysts said Oracle’s fundraising plan helped answer the “how do you pay for it?” question — at least for now. Guggenheim said the plan sends “a clear message to bond investors and the rating agencies,” while Barclays said the mix of equity and mandatory convertibles should calm debt markets and strengthen the balance sheet; Jefferies called it a plan that “buys time” but warned it could pressure margins and delay positive free cash flow until fiscal 2029. (Reuters)
OpenAI has also tried to tamp down the idea that its supply chain politics are turning into a feud. OpenAI chief Sam Altman wrote on X: “We love working with NVIDIA,” adding he did not “get where all this insanity is coming from,” Business Insider reported. Oracle also posted that “The NVIDIA-OpenAI deal has zero impact” on its OpenAI relationship and said it remained confident OpenAI can raise funds and meet commitments, the report said. (Business Insider)
The swirl is starting to look less like a personality story and more like a financing story. Quartz pointed to a Wedbush note urging investors to “see forest through the trees,” calling the Nvidia-OpenAI back-and-forth a “mini soap opera” as markets begin treating AI like infrastructure: slow to build, expensive to fund, and sensitive to credit conditions. (Quartz)
Still, the downside case is easy to sketch. If OpenAI and other big buyers shift more inference work to alternative chips faster than expected, Nvidia could face pressure at the edge of its business even if training demand holds up. For Oracle, the plan implies dilution risk and more debt at a time investors are already jumpy about whether AI spending will produce steady cash returns.
For now, markets are left with two overlapping questions: how quickly AI usage grows, and who finances the compute that growth requires. The answers will come less from keynote stages than from term sheets, chip road maps and bond calendars.
