Oracle faces $7 billion guarantee for AI data center energy

What happened?

Oracle, in collaboration with Vantage and OpenAI, is building the Lighthouse Campus, a nearly one-gigawatt data center in Port Washington, Wisconsin, valued at around $15 billion. According to an affidavit filed by Oracle with the local regulator, the company must post a $7 billion guarantee to ensure it pays the electricity bill during construction and initial operation. This guarantee, demanded by the local utility, is one of the largest ever requested for such a project. The amount is roughly 47% of the campus's total cost, reflecting the magnitude of risk perceived by the utility. The project, expected to be operational in phases starting in 2027, will consume as much electricity as a city of 200,000 people. The affidavit, obtained by The Next Web, reveals that Oracle has already deposited $1.5 billion in cash as part of the guarantee, with the remainder covered by letters of credit and other financial instruments. The local utility, We Energies, justified the demand by arguing it needs guarantees to cover the costs of expanding the power grid, including new substations and transmission lines, with an estimated additional cost of $3.5 billion.

Why is this important?

This case illustrates the energy bottleneck facing the AI industry. The electricity demand of AI data centers is growing exponentially: according to the International Energy Agency (IEA), data center electricity consumption could double by 2026, reaching 1,000 TWh, equivalent to Japan's total consumption. Companies must compete for limited resources, and the demand for such a large guarantee reflects the risk utilities perceive from large-scale projects with unprecedented energy consumption. Moreover, it sets a legal and financial precedent: other tech companies could face similar conditions in future expansions. For example, in 2024, Google faced delays on a data center in Finland due to lack of local grid capacity and had to invest in electrical infrastructure upgrades. This case could lead utilities worldwide to demand higher guarantees, especially in regions with already strained power grids, such as Northern Virginia (Data Center Alley) or Singapore, where moratoriums on new data centers have been imposed.

Consequences for the sector

• Barriers to entry: Only companies with massive cash reserves will be able to afford these upfront costs, potentially consolidating the dominance of giants like Oracle, Microsoft, and Google. AI startups needing computing capacity may be forced to rent space in existing data centers, paying hefty premiums. According to a 2025 McKinsey report, the cost of capital for new data centers has increased by 30% over the past two years due to energy guarantee requirements.
• Impact on rates: The costs of guarantees and electrical infrastructure investments could be passed on to consumers and businesses using cloud services. We Energies has already requested a 12% rate increase for residential customers to cover grid upgrades, according to regulatory documents. In the long term, cloud computing costs could rise by 5% to 10% annually, according to Gartner estimates.
• Innovation in efficiency: The search for more efficient cooling technologies, renewable energy, and alternative locations (e.g., near nuclear or hydroelectric plants) will accelerate. Oracle has already announced plans to use liquid cooling at Lighthouse Campus, reducing water and energy consumption by 20% compared to traditional systems. Additionally, companies like Microsoft are exploring modular data centers that can be rapidly deployed near renewable energy sources.
• Regulatory pressure: Local and state governments may review their incentive and permitting policies to balance economic development with energy sustainability. Wisconsin, for example, has granted Oracle tax breaks worth $200 million but now faces criticism from environmental groups who say the data center will increase the region's carbon emissions. In the European Union, the 2024 Energy Efficiency Directive requires data centers to report their energy consumption and establish reduction plans, which could be further tightened.

What should readers know?

This case is not isolated. In 2024, Amazon also faced delays due to lack of power at its data centers in Virginia and had to pay $500 million to secure additional electrical capacity. The difference here is the magnitude of the guarantee, which exceeds any known precedent. Previously, the largest energy guarantee for a data center was $2.5 billion, demanded from Meta in 2023 for a project in Louisiana. Readers should understand that AI infrastructure requires not only powerful servers but also massive and stable electricity supply, which is creating tensions with local power grids and raising entry costs. For investors, it signals that data center margins could compress, especially if energy costs continue to rise. For users, it could translate into higher cloud service prices or a slowdown in AI expansion, as companies may prioritize efficiency over deployment speed. Additionally, the case could drive greater investment in renewable energy and energy storage, as well as more efficient AI technologies like specialized chips (TPUs, NPUs) that consume less electricity per operation.

Historical context

In the 1990s, building data centers for the internet bubble also faced energy constraints, but the current scale is much larger. A single AI data center can consume as much electricity as a small town. For example, OpenAI's data center in Iowa consumes 150 MW, equivalent to 30,000 homes. The difference now is that demand is concentrated among a few companies, giving them bargaining power but also exposing them to significant regulatory and financial risks. During the dot-com bubble, many companies built speculative data centers that later sat empty, but today's demand is backed by real revenue from cloud and AI services. However, the risk of overcapacity exists if AI growth slows. In 2025, consulting firm IDC estimated that AI data center utilization rates are 65%, suggesting room for more capacity, but energy costs could limit profitability. Another historical parallel is with aluminum smelters in the 20th century, which required large amounts of electricity and were often located near hydroelectric dams. Similarly, AI data centers are seeking locations with access to cheap energy, such as Washington state (hydroelectric) or Texas (wind and solar).

“The demand for a $7 billion guarantee is a sign that the energy market is not prepared for AI demand. Tech companies will have to innovate not only in software but also in how they obtain and finance their electricity.”

In summary, the Oracle Lighthouse Campus case is a turning point that underscores the critical intersection between AI expansion and energy infrastructure. Tech companies, regulators, and investors will need to collaborate to prevent energy bottlenecks from slowing AI progress while ensuring a sustainable and affordable supply for all.