California Makes AI Data Centres Pay for Their Own Power and Water as Newsom Signs Seven-Bill Package

California Governor Gavin Newsom has signed a package of seven bills that require large data centres to disclose their electricity and water use, pay for upgrades to the grid and water systems they rely on, and face tighter local review, in what is widely seen as the most aggressive response yet by a US state to the physical footprint of the artificial-intelligence boom.

The laws, reported by Reuters on 22 September 2026, direct the California Public Utilities Commission to create a separate rate class for large data centres, so that transmission, generation and wildfire-related costs associated with serving them are not shifted onto households. Developers will also be required to report projected water consumption before receiving local approval, and projects will lose certain shortcuts around environmental review.

Newsom said communities should remain in the driver's seat, and that companies profiting from AI infrastructure should not do so at residents' expense.

Why data centres became a political issue

For most of the internet era, data centres were largely invisible to the public, tucked away in industrial parks and rarely the subject of political debate. The AI boom has changed that. Training and running large AI models requires enormous computing power, and hyperscale technology companies have been racing to secure land, electricity and water for new campuses.

Individual AI data-centre projects can require hundreds of megawatts of power, comparable to the demand of a small city. Utilities across the United States have reported surging requests for new connections from data-centre developers, forcing them to plan new generation and transmission investments.

Those investments are typically paid for by all utility customers through electricity rates. As a result, households and small businesses have raised concerns that they could end up subsidising infrastructure built primarily to serve large technology companies. In California, where electricity rates are already among the highest in the country, the issue has particular resonance.

What the new laws do

The seven-bill package addresses several aspects of data-centre development. The creation of a separate rate class is designed to ensure that the costs of serving large data centres are borne by those facilities rather than spread across residential customers. That includes costs related to transmission, generation and wildfire mitigation, a significant and growing expense for California utilities.

Disclosure requirements will provide regulators and the public with data on how much electricity and water data centres consume, information that has often been difficult to obtain. Requiring developers to report projected water consumption before local approval gives communities a clearer basis for evaluating proposals, particularly in drought-prone areas.

The changes to environmental review mean data-centre projects will no longer benefit from certain streamlined pathways, subjecting them to more thorough assessment of their environmental impacts. Local governments will gain more scrutiny over projects, potentially slowing approvals but increasing accountability.

Water in a drought-prone state

Water use is a particularly sensitive issue in California, which has experienced repeated severe droughts. Many data centres use water for cooling, either directly through evaporative systems or indirectly through the electricity they consume. The amount varies widely depending on design and location, and some operators have adopted technologies that reduce water use.

By requiring disclosure of projected water consumption, the new laws aim to give communities and water agencies the information needed to plan for competing demands among agriculture, households, industry and ecosystems.

The industry's perspective

The technology industry has argued that data centres are essential infrastructure for the digital economy and for AI innovation, and that burdensome regulation could push investment to other states or countries. Industry groups have also pointed to efforts by operators to procure renewable energy, improve efficiency and reduce water use.

The new laws are likely to raise the all-in cost of building data centres in California. Operators will need to fund their own grid connections, publish usage data and navigate more rigorous review processes. Some may choose to locate new campuses in states with lower electricity costs and lighter regulation, such as Texas or parts of the Midwest, or overseas.

Startups that rent computing capacity from cloud providers may not pay these costs directly, but they could feel them indirectly through higher compute prices if California sites become slower or more expensive to develop.

A model for other jurisdictions

California's approach is likely to be closely studied elsewhere. Several US states have begun debating how to manage data-centre growth, and Texas has already legislated on how very large electricity loads connect to its grid. Internationally, countries including Ireland and the Netherlands have grappled with the strain data centres place on electricity grids, and in some cases have restricted new connections in congested areas.

In the United Kingdom, where data centres have been designated critical national infrastructure, the debate has focused on balancing AI ambitions with local opposition over energy and water. Governments seeking to attract AI investment face a difficult trade-off between economic benefits and the costs borne by communities.

Implications for India

For India, which is experiencing its own data-centre boom driven by cloud adoption, data-localisation requirements and AI ambitions, California's move offers useful lessons. Indian states are competing to attract data-centre investment with incentives on land, power and approvals. As capacity grows, questions about electricity tariffs, renewable-energy sourcing and water use in water-stressed cities such as Mumbai, Chennai and Hyderabad will become more pressing.

Policymakers may need to consider how to ensure that data-centre growth supports, rather than strains, local infrastructure and communities, and how to allocate the costs of grid upgrades fairly.

Enforcement and implementation will now shift to regulators and local governments. The Public Utilities Commission will need to design the new rate class, determining how costs are allocated and how quickly data centres must pay for upgrades. Local authorities will need to incorporate new disclosure requirements into their planning processes. The details of these rules will determine how significantly the legislation reshapes data-centre development in the state.

The broader significance

California's legislation marks a turning point in how governments approach the AI boom. The debate is shifting from the capabilities of AI models to the physical infrastructure required to run them: land, power, water and grid capacity. By turning data-centre power and water from a local zoning issue into statewide cost and disclosure rules, California is asserting that the benefits of AI must be weighed against its resource demands.

The laws will not stop data-centre development in California, but they will change its economics and increase transparency. As other jurisdictions consider their own responses, California's experiment could shape the global conversation about how to build the infrastructure of the AI age sustainably and fairly.