The artificial intelligence revolution has inadvertently done what decades of fossil fuel advocacy could not: ignite the biggest construction boom in natural gas-fired power plants in history. According to recent reporting, the insatiable electricity appetite of AI data centers is driving utilities to build new gas plants at an unprecedented pace, while also extending the life of aging coal facilities that were slated for retirement.
The arithmetic is stark: a single large data center can consume as much electricity as a mid-size city, and the speed at which these facilities are being built far outpaces the construction timelines of renewable energy sources like wind and solar. This has created a tension point between the tech sector's expansion and the urgent need to decarbonize the global economy.
Why AI needs so much power
Modern AI models require massive computational resources. Training large language models and running inference operations involves thousands of specialized chips, often GPUs, operating in parallel within enormous server clusters. These servers generate heat and require intensive cooling, further boosting energy consumption. The International Energy Agency has noted that data centers could consume up to 1,000 terawatt-hours of electricity by 2026, roughly equivalent to the total electricity consumption of Japan. This demand is straining grids that were already under pressure from electrification of transport and heating.
The explosion of generative AI tools since late 2022 has accelerated the build-out of data centers. Companies like Microsoft, Google, Amazon, and Meta are racing to expand their cloud and AI infrastructure, often committing billions of dollars to new facilities. But the energy infrastructure to support them is not green enough to meet existing climate pledges. Utilities, facing reliability concerns, are turning to the fastest dispatchable source: natural gas.
The states drawing lines
Several U.S. states are attempting to force the issue through legislation. A bill currently on New York Governor Kathy Hochul's desk would require large data centers to hit renewable energy benchmarks starting in 2030, reaching at least 90% renewable energy by 2040. The bill's author, state Senator Kristen Gonzalez, argues that the targets are achievable given the immense financial resources of tech giants. These are the wealthiest companies on earth, she has said, and they can afford to build the clean power they need.
Michigan, Oregon, and Minnesota have already passed laws in the last 18 months to protect existing commitments to emissions-free electricity by 2040. Michigan's approach ties the requirement to tax incentives: hyperscale data centers must reach 90% clean energy within six years to keep a lucrative sales tax exemption. Similar bills have been introduced in California, Illinois, New Jersey, Pennsylvania, and Virginia.
These legislative efforts represent a quiet but determined fight to prevent the AI boom from derailing climate goals. Environmental groups and clean energy advocates argue that without such laws, the data center build-out will lock in decades of fossil fuel dependence.
An honest admission from the other side
Perhaps the most telling quote in the original reporting comes from Bob Jenks of the Oregon Citizens' Utility Board. He conceded that the 2040 clean energy target is hard to meet with data centers, and hard to meet without them. This captures the fundamental dilemma: the clean energy goal was already ambitious, and AI has pulled it further out of reach. Households are feeling the impact first, with electricity bills climbing across many utility territories. AI data centers are also driving up power costs at Rust Belt factories, where industrial consumers face higher rates as utilities shift costs onto remaining customers.
The regulatory back door
Unable to outbuild the boom, advocates have turned to regulatory tactics. The emerging strategy is to get state regulators to allow large power users to build their own clean generation and plug it directly into the grid. In Colorado, regulators ordered Xcel Energy to create such a program. In an April filing, Xcel accepted that the program could benefit customers, citing Google projects connecting 115 megawatts of geothermal in Nevada and 1,900 megawatts of wind, solar, and storage in Minnesota.
Google's deal with NV Energy is considered the first of its kind, and the company says similar arrangements are approved or pending in eight more states. The Corporate Energy Buyers Association (CEBA) struck a comparable deal with Georgia Power and is now working on a similar agreement in North Carolina. The pitch to utilities is commercial: they gain a huge long-term customer who pays to expand the grid, rather than watching that customer build standalone generation and leave the utility system.
Why this is the real fight
Grid access is where the outcome will be decided, not in the legislature. Regulators have been fast-tracking data center grid connections, and whoever controls that queue controls what gets built. The interconnection process determines not only where data centers are located but also what kind of power they draw from the grid. Money is chasing this bottleneck, with Nvidia-backed startups raising capital to solve data center power issues. Energy, not silicon, has become the binding constraint on AI.
Communities are pushing back independently. In a single quarter, local opposition blocked 75 data center projects worth $130 billion, citing concerns over noise, water usage, and the burden on local grids. Congress is circling too, with the House voting on a bill to push data center energy costs back onto the companies creating them.
According to CEBA's policy chief, the decisions being made now will set energy policy for two or three decades. That assessment is probably accurate, which is why the technical argument about grid interconnection deserves far more attention than it typically receives. The gas plants are being poured in concrete while the rules are still being written, and concrete tends to win those races.
Meanwhile, the broader context is that the US power grid is already struggling to integrate renewable energy at the scale needed to meet climate targets. The additional demand from AI comes at a moment when many coal plants are being retired, and nuclear plants are aging. Natural gas, with its lower carbon emissions than coal but still significant methane leakage, is seen as a bridge fuel. But the bridge is growing longer, and environmentalists fear that the AI boom will make it permanent.
The stakes extend beyond the United States. Europe and Asia are also grappling with data center energy demands. In Ireland, data centers accounted for 21% of all metered electricity in 2023, forcing regulators to pause new connections near Dublin. In Singapore, a moratorium on new data centers was lifted in 2022 only after operators agreed to adopt higher energy efficiency standards. The global race to host AI infrastructure is creating similar tensions between economic development and climate commitments.
The quiet fight against the gas plant boom is not just about emissions. It is about the shape of the future grid. If data centers can be powered by clean energy through innovative tariff structures and on-site generation, they could become an engine for renewable deployment. If they continue to lean on fossil fuels, they will lock in emissions for decades. The outcome depends on whether regulators, lawmakers, and tech companies can align incentives before the concrete sets.