Why is the Oracle Wisconsin AI campus facing power delays?
The primary cause of the delays is the restart of the regulatory review process required to connect the massive Project Lighthouse campus to the local electrical grid. Although the Public Service Commission of Wisconsin (PSC) had initially found the application from the American Transmission Company (ATC) to be complete in December 2025, subsequent changes to the project forced a total reset of the legal timeline.
Between January and July 2026, ATC submitted 564 additional documents that fundamentally altered the scope of the transmission project. These modifications included changes to transmission routes, the introduction of temporary bypass lines, and updated cost estimates. Because the scale and nature of these changes were not clearly itemized, the PSC withdrew its completeness finding in August 2026, effectively nullifying ten months of prior review work.
The regulatory uncertainty stems from the difficulty in assessing the environmental and technical impact of these late-stage additions. An administrative law judge had ordered ATC to provide a detailed breakdown of every alteration on July 1, 2026, but found by July 15 that the company had failed to comply with that order. Consequently, the PSC closed the case on September 10, 2026, without making a decision on the merits, leading to a refiling by ATC on September 18, 2026, which restarted the statutory clock from zero.
How is physical construction progressing despite grid uncertainty?
Physical construction at the Port Washington site is currently moving ahead of the electrical infrastructure timeline, creating a situation where buildings may be completed long before they can be powered. Site imagery from September 2026 confirms that two buildings have already been enclosed, and equipment installation is already underway in certain completed sections of the campus.
Work on the campus has been continuous since December 2025, supported by existing permits. However, the gap between structural readiness and electrical connectivity poses a significant logistical challenge for Oracle. As Javier Reyes, Head of Data Center Research at Aterio, noted, while construction is progressing steadily, the actual delivery of power is entirely dependent on a regulatory process that has been forced back to the beginning.
The disconnect between building completion and operational readiness
The discrepancy between the physical state of the campus and its operational capacity is a growing concern for stakeholders. While the shells of the data centers are rising, the lack of a finalized transmission plan means the site remains a collection of high-tech buildings without a reliable energy source. This creates a "stranded asset" risk where capital is tied up in infrastructure that cannot yet generate revenue or support AI workloads.
What are the specific power requirements for Project Lighthouse?
There is a notable difference between the reported IT demand and the total electrical requirements of the Wisconsin campus, a distinction that is critical for understanding the project's true scale. Vantage, the developer for Oracle, describes the campus as having an IT demand of 902 MW, whereas utility records indicate a total requirement of 1.3 GW.
This gap of approximately 400 MW is attributed to the energy required for auxiliary systems. The 1.3 GW figure accounts for the total electrical load, which includes the electricity needed for cooling systems and other essential facility operations that support the computing hardware. For a facility of this magnitude, cooling is a massive energy consumer, making the distinction between IT load and total facility load vital for grid planning.
| Metric Type | Capacity (MW/GW) | Description |
|---|---|---|
| IT Demand | 902 MW | Direct power consumed by computing equipment. |
| Total Electrical Requirement | 1.3 GW | Total load including cooling and facility operations. |
| Initial 2027 Request | ~310 MW | Requested partial service during the first year. |
| 2027 Peak Demand | 438 MW | Projected peak load for the initial phase. |
What is the impact of these delays on Oracle's delivery timeline?
The regulatory reset introduces a "meaningful risk" that the power required for the campus will arrive much later than Oracle has previously indicated to its customers. The current trajectory suggests that the original goal of providing significant capacity in 2027 is no longer viable under the current regulatory timeline.
If the regulatory review continues to face delays, the availability of power will shift significantly. Instead of receiving partial service in 2027, the campus might not see partial grid access until June 2028. Furthermore, the full 1.3 GW capacity required to run the campus at its intended scale would likely be pushed to April 2029. This timeline is well beyond the guidance Oracle has provided to its clients, potentially impacting the deployment of large-scale AI models that rely on this specific infrastructure.
The delay is not merely a matter of months but a fundamental shift in the project's lifecycle. As the statutory clock has restarted, the window for environmental reviews and public hearings must be reopened, adding layers of complexity that were not present in the initial application. For Oracle, this means the massive investment in Wisconsin may not yield the expected computational throughput for several years longer than anticipated.
FAQ: Oracle Wisconsin AI campus delays
Why did the PSC withdraw the completeness finding for the ATC application?
The Public Service Commission withdrew the finding because the American Transmission Company (ATC) failed to clearly itemize 564 new documents and project alterations. These changes, including new routes and cost updates, made it impossible for regulators to determine the exact scope of the project under review.
What is the difference between 902 MW and 1.3 GW for this project?
The 902 MW figure represents the IT demand, which is the power used directly by computing hardware. The 1.3 GW figure is the total electrical requirement, which includes the additional power needed for cooling systems and other essential facility operations.
Will construction stop because of the power delays?
No, construction at the Port Washington site is continuing steadily. Permits have been secured, and buildings are already being enclosed and fitted with equipment. The delay specifically affects the delivery of electricity via the transmission lines, not the physical building process.
When could the campus be fully operational?
Due to the regulatory restart, full 1.3 GW availability may be delayed until April 2029. This is a significant shift from previous expectations, as the current regulatory process has pushed the timeline well beyond Oracle's initial guidance.
Who is developing the Project Lighthouse campus?
The campus, known as Project Lighthouse, is being developed by Vantage for Oracle. It is located in Port Washington, Wisconsin, and is designed to house four major data centers to support AI workloads.
Key takeaways
- The regulatory review for the 1.3 GW Wisconsin AI campus has restarted from zero.
- ATC's filing of 564 additional documents caused the PSC to withdraw its completeness finding.
- Physical construction is advancing, with two buildings already enclosed as of September 2026.
- Full 1.3 GW power availability may be delayed until April 2029.
- Total electrical demand is 1.3 GW, while IT-specific demand is 902 MW.
The future of Project Lighthouse and AI infrastructure
The delays facing Project Lighthouse highlight a growing tension between the rapid expansion of AI computing needs and the slower pace of electrical grid modernization. While Oracle and Vantage have moved quickly to build the physical infrastructure, the regulatory and utility frameworks are struggling to keep up with the massive power requirements of next-generation data centers. The success of this campus will ultimately depend on whether the American Transmission Company can navigate the renewed regulatory scrutiny to deliver the necessary high-voltage lines. For now, the project serves as a cautionary tale regarding the complexity of scaling AI infrastructure at the gigawatt level.
