Key Takeaways:
Oracle's $90 billion AI infrastructure bet is running into a wall of state-level regulatory pushback and a credit downgrade that could add billions in financing costs to its biggest projects.
Oracle plans to spend $90 billion on data center construction in fiscal 2027, up 70% from $55.7 billion in fiscal 2026, according to company filings. The spending spree — funded partly through $40 billion in new debt and equity — is meant to cement Oracle's position as a primary compute provider for OpenAI under a five-year contract valued at $300 billion starting in 2027. But the company's credit profile is deteriorating faster than its buildout.
S&P Global downgraded Oracle to BBB-minus in recent months, one notch above junk status, citing that "growing AI infrastructure business is diluting its strong business risk profile." The downgrade triggered an unexpected consequence: the Public Service Commission of Wisconsin refused to waive financial collateral requirements for Oracle's planned 1-gigawatt data center in the state, demanding more than $7 billion in cash or letters of credit simply to connect to the power grid.
"The financing costs would impact future investment in the state," Oracle argued in its petition against the requirement, according to regulatory filings. The commission countered that existing ratepayers should not subsidize data center construction, a position gaining traction in statehouses nationwide.
State Legislatures Push Back on Data Center Boom
More than 40 states are now imposing various regulations on data center development, with 15 states considering outright bans, according to legislative tracking data. Lawmakers in Wisconsin, New York and Maine have led the charge, citing concerns about strained power grids, rising electricity costs and water consumption. The pushback represents a risk Oracle cannot solve through cost-cutting alone.
The company has already taken extreme measures to preserve cash. Oracle cut 21,000 jobs by the end of fiscal 2026, reducing its workforce to 141,000 from 162,000, in a move that could free up to $10 billion annually. But the savings are a fraction of what's needed. Oracle's cash flow swung from positive $11.8 billion in fiscal 2024 to negative $23.7 billion in fiscal 2026, a deterioration of more than $35 billion in two years.
The Wisconsin collateral requirement alone would cost roughly $100 million annually in ongoing maintenance fees, according to regulatory estimates, on top of the project's $15 billion total price tag. If other states follow Wisconsin's lead, Oracle could face similar demands across its planned data center portfolio.
OpenAI Dependency Adds Another Layer of Risk
Oracle's balance sheet is increasingly tied to the success of one customer. Of the $638 billion in future contract volume on Oracle's books, about 50% comes from OpenAI alone, according to SEC filings. S&P has described OpenAI as a "central credit risk" for Oracle.
Analyst David Desjardins warned that Oracle's entire strategy hinges on OpenAI's ability to keep raising capital. "If OpenAI faces any hurdles in raising money, the situation could drive Oracle towards the eye of the storm," he said. OpenAI relies on capital markets remaining open to meet its rising financial obligations, and any disruption could cascade directly to Oracle's revenue pipeline.
Oracle shares have fallen 35% year-to-date in 2026 and 47% over the past 12 months, reflecting investor concern about the widening gap between capital commitments and visible returns. The company trades at roughly 18 times forward earnings, a discount to Microsoft at 28 times and Amazon at 32 times, as the market prices in execution risk.
For hyperscaler competitors Microsoft and Amazon, the regulatory headwinds are manageable given their diversified revenue streams and investment-grade balance sheets. For Oracle, which is spending nearly 13 times its fiscal 2024 capex of $6.9 billion, the margin for error is far thinner. The next 12 months will determine whether state regulators force a recalibration of the industry's most aggressive AI infrastructure buildout.
This article is for informational purposes only and does not constitute investment advice.
