Oracle OCI: AI Infrastructure Power Bottleneck and Financing Risk, October 2026

Oracle shares closed at $130.29, down 9%, on the October 13 session, with no single report explaining the move; the backdrop includes a $55.66B FY2026 capital-expenditure programme equal to 83% of revenue — the highest intensity among tracked hyperscalers — and accumulating challenges around power supply, construction timelines, and financing structure.

Oracle shares closed at $130.29 on the October 13 session, a loss of 9%. No coverage item reviewed by Slicast explicitly identifies what drove that session's decline. The move arrives within a broader October context that has been anything but quiet: International Business Times reported that AI credit concerns were specifically weighing on Oracle and Broadcom, even as NVIDIA proved comparatively resilient; and earlier in the month, a reported revision in OpenAI's annualised revenue figure — down to roughly $50 billion from the approximately $70 billion that had reportedly helped lift Oracle shares, with October reporting noting the stock was then poised for its best weekly performance since 1999 — contributed to a sector-wide selloff on October 8 that swept in Oracle, NVIDIA, CoreWeave, and other AI-infrastructure names. Whether those dynamics extended into the October 13 session remains unconfirmed by available reporting.

The structural backdrop, however, is unambiguous. Capital expenditure reached $55.66 billion in FY2026, a 162.4% increase from $21.21 billion in FY2025, according to Slicast's compilation of SEC XBRL filings — a dataset covering 66 reporting periods from 2010 through 2026. At 83% of FY2026 revenue of $67.36 billion, Oracle's capex intensity is the highest among the seven hyperscalers tracked by Slicast; in absolute terms the company ranks fifth, with Amazon leading the group. The single quarter ending August 31, 2026 alone accounted for $28.50 billion in capital spending, the largest single-quarter figure on Slicast's record for the company. To sustain this pace, Oracle has reportedly joined Broadcom and SpaceX in seeking additional Wall Street financing for AI compute purchases; analysts have observed that the leading hyperscalers are collectively targeting more than $1 trillion in combined capital expenditure, outrunning their operating cash flows and increasingly shifting infrastructure liabilities into off-balance-sheet vehicles as credit markets intensify scrutiny of the sector's spending rate.

The most acute near-term constraint is energy, not demand. Oracle invoked force majeure on Project Jupiter, its AI data centre development in New Mexico, citing power-infrastructure obstacles — a situation first reported in September alongside accounts of rent-payment delays at the same site. In October, Oracle proposed up to $55 million in annual power bill credits and reaffirmed Bloom Energy's contractual role at the project, signalling active renegotiation rather than exit; Bloom Energy's CEO publicly dismissed concerns about the contract unravelling, describing the company's fuel-cell units as interchangeable within Oracle's broader energy architecture. Separately, Oracle has reportedly been trucking natural gas directly to data centres where pipeline connections remain delayed. At the Wisconsin campus — planned at 1.3 gigawatts — regulatory approval for grid interconnection has been restarted and the baseline timeline for full power has been extended to October 2028, a delay that analyst firm Aterio had flagged as a material risk. IO Fund has argued that for Oracle and comparable operators, power, not demand, is now the defining bottleneck. Oracle has also reportedly secured a nuclear power agreement to provide firm baseload electricity for its data centre portfolio over the longer term, while a separate $18 billion facility reportedly survived a significant flooding scare, illustrating the breadth of physical-infrastructure risks accumulating at scale.

Against these execution challenges, demand-side signals remain substantive. Tencent has reportedly agreed to lease 100,000 AI chips from Oracle's Southeast Asia data centres in a five-year arrangement valued at roughly $7 billion — a transaction corroborated across multiple reports and structured, according to those accounts, to give Tencent access to GPU capacity that US export controls render unavailable on the Chinese mainland. Oracle's Fusion Claw inference platform has been reported to reduce enterprise AI compute costs, potentially widening the addressable customer base and differentiating OCI in a market where Amazon, Microsoft, and Google Cloud hold significant scale advantages. Goldman Sachs has projected that major technology companies could collectively spend $1.2 trillion on AI infrastructure through 2027, a forecast that, if realised, would support Oracle's argument that the demand runway justifies the scale of capital commitment.

The investment thesis on Oracle is being stress-tested on two simultaneous fronts. The bull case is structural: enterprises need GPU-dense cloud capacity, hyperscalers need infrastructure partners willing to build at scale, and OCI has repositioned itself as an aggressive alternative to AWS, Azure, and Google Cloud by deploying capital at intensity ratios its larger rivals have not replicated. The bear case is equally concrete: an 83% capex-to-revenue ratio leaves almost no tolerance for slippage, and the accumulating record of force majeure notices, grid-approval setbacks, improvised energy logistics, and physical-site risks suggests that execution challenges are systemic rather than episodic. Off-balance-sheet financing structures, if expanding in scope, add a layer of credit opacity that markets may be beginning to price — without, as yet, any single published report pinning that concern to the October 13 session specifically. Three signals warrant close monitoring ahead: whether the Wisconsin 1.3-gigawatt campus achieves grid-interconnection milestones materially ahead of its revised October 2028 baseline; how credit markets and rating agencies respond to Oracle's reported push for additional AI-compute financing; and whether GPU-leasing contracts — with the Tencent arrangement as the most visible example — generate recurring revenue growth at a rate sufficient to close the gap between Oracle's capital deployment pace and its current revenue base.