Ellison doubles down on AI infrastructure as investors worry about debt-fueled boom

Larry Ellison’s Oracle is spending big on AI infrastructure — and investors are increasingly alarmed.

The New York Times reports that Oracle and partners, including OpenAI, planned up to $500 billion in spending on Project Stargate over four years, aiming to deploy roughly 10 gigawatts of computing capacity; the Times also says OpenAI agreed to pay Oracle about $300 billion over roughly five years beginning in 2027, figures the company disputes in public filings and statements The New York Times. Reuters separately reported Oracle expects to raise $45 billion to $50 billion in 2026 through debt and equity to build cloud capacity, and cited bondholder lawsuits over disclosures tied to those financing plans Reuters.

Why it matters: Ellison’s bet has recast Oracle from a legacy enterprise software vendor into a capital-intensive AI infrastructure contender. The strategy briefly inflated Oracle stock and Ellison’s net worth; the Times says shares rose 43% on the Stargate narrative and later fell as much as 60% from the peak, turning gains into a test of whether AI demand will justify massive, debt-funded buildouts The New York Times.

The $500 billion Stargate claim and the gaps in public proof

The Times frames Project Stargate as potentially the largest AI infrastructure program in history, citing a planning figure of up to $500 billion and a goal of 10 gigawatts of compute. Those headline numbers come from people briefed on the plans; Oracle’s regulatory filings and public statements do not provide a line-item confirmation that matches that scale. The company has acknowledged expanding data‑center capacity for cloud and AI, but detailed cost and customer‑commitment figures remain opaque, leaving analysts to infer scale from hiring, land purchases and long‑term contracts described in the Times reporting The New York Times.

That opacity is the central complaint of critics. Reuters reports bondholders have sued, alleging Oracle failed to disclose financing needs tied to AI infrastructure, and credit agencies have pushed the company down the ratings ladder toward speculative territory Reuters. The suits and downgrades underscore that Oracle’s ambition is being financed largely through markets rather than operating cash flow.

How Oracle’s wager compares with Microsoft, AWS and Google Cloud

Oracle’s approach contrasts with rivals that have scaled more gradually or diversified risk. Microsoft leverages a broad Azure platform and an existing strategic tie to OpenAI without concentrating financing on a single hyperscaler play, while Amazon Web Services benefits from the deepest installed enterprise demand and cash flows among clouds. Google combines internal AI research with its data‑center footprint, giving it another path to scale without the same headline debt buildup. The Times frames Oracle as uniquely concentrated on bespoke AI deals and large data‑center expansion — a posture that magnified stock gains when the market believed in Stargate and magnified losses as skepticism grew The New York Times.

Investors and some analysts question the scaling assumption behind the bet: that more raw compute will produce a reliably pro rata revenue stream. The Times says skeptics worry returns on gargantuan infrastructure are unproven; Reuters’ coverage of debt taps during the AI spending surge adds a concrete counterpoint — companies are funding capacity now, before enterprise consumption patterns have normalized Reuters.

Oracle disputes some of the reporting and continues to tout long-term cloud contracts. Yet the combination of aggressive market proclamations, lawsuits from bondholders and a credit downgrade has turned Ellison’s bet into a cautionary tale inside the sector: a reminder that infrastructure scale requires not just engineering but disciplined finance.

The next events to watch are Oracle’s 2026 financing plans and bondholder litigation timelines, and whether confirmed customer contracts materialize to shift cash‑flow projections — concrete signals that will determine if Ellison’s wager becomes a defining victory or the face of an overheated AI bubble.