Oracle's 1.3GW Wisconsin Data Center Stalled by Grid Approval, Full Power Supply Potentially Delayed to 2029
nashnova research
Oracle's 1.3GW mega data center in Wisconsin has been forced back to square one after regulators revoked the grid permit's completeness finding. Base-case full power is now October 2028; if approvals slip again, April 2029 — with each month of delay deferring roughly $850–900 million in revenue.
Why did the approval process reset to zero?
Transmission company ATC had its application declared "complete" in December 2025, but then filed 564 supplemental documents, adding new route options and a temporary bypass line.
Wisconsin's Public Service Commission (PSC) unanimously revoked the completeness finding — reportedly the first time in decades, possibly ever.
In plain terms = the application changed so many times that regulators ruled it was no longer the same application and sent it back to the starting line.
The PSC closed the original case on September 10. ATC refiled on September 18; the statutory review clock restarted from zero.
How long is the delay? What do the three scenarios look like?
Research firm Aterio mapped Wisconsin's statutory timeline to three scenarios: optimistic full power by August 2028, base case October 2028, delayed April 2029.
This means → Oracle's original commitment to deliver capacity to tenants by H2 2027 is effectively dead — at least a year late.
The key detail: historically the PSC has invoked an extension in every comparable large-transmission case, with actual review times running 355–358 days versus the statutory 180.
In plain terms = the nominally fastest "optimistic" scenario has never actually happened; the delayed scenario is the one with historical precedent.
Why did costs jump 75%?
ATC's refiled transmission project now carries an estimated cost of $2.48–2.72 billion, up roughly 75% from the original $1.3–1.7 billion.
About $1.1 billion is earmarked for grid-stability systems; Oracle has committed to covering that portion.
This means → even if the project is eventually built, Oracle's upfront outlay far exceeds the original budget, stretching the payback period.
How serious are the financing and credit pressures?
Construction funding for the Wisconsin campus comes from $15 billion carved out of a Vantage $38 billion loan package that reportedly struggled in syndication.
Wisconsin's large-load customer tariff requires a credit rating of A- or above; Oracle is currently rated BBB- — several notches short.
This means → Oracle must post additional letters of credit. The company itself disclosed the annual cost could exceed $100 million, with total guarantees reportedly topping $7 billion.
What is the financial hit from the delay?
Goldman Sachs estimates hyperscalers need roughly $11.6 billion in annual AI revenue per GW of compute to earn a 15% return on 2026–2027 AI capex.
Lighthouse's IT load is about 0.9 GW. Each month of delay defers roughly $850–900 million in revenue.
Hardware capex is typically spent two to three months before an asset goes live — what gets delayed is revenue, not cost.
Oracle's FY27 capex guidance runs as high as $95 billion, up sharply from $55.7 billion in FY26, further squeezing financial flexibility.
Is this still an "isolated incident"?
Wisconsin is the second of Oracle's five core data-center campuses to be flagged for delay risk.
Just two weeks earlier, Barclays told clients the New Mexico issue had "no contagion to other projects" — Wisconsin is disproving that claim in real time.
This reflects a systemic pattern: New Mexico is stuck on pipeline and air permits; Wisconsin is stuck on grid approval. The causes differ, but the result is the same — assets are not coming online on schedule.
In plain terms = the logic behind hundreds of billions in debt-funded construction is that projects deliver revenue on time. When every project has "its own unique reason for delay," the "isolated incident" defense no longer holds.
市场有风险，内容仅供研究参考，不构成投资建议。