AI is already the most transformative technology of our lifetime. With that comes a healthy dose of unknowns and difficult questions in search of answers. Does transformation ultimately lead us somewhere better or worse? Ask some in the climate community about AI, and they’ll say the emissions associated with energy demand growth point to a climate catastrophe.
I don’t believe they do.
The truth is that while AI is poised to reshape every aspect of the global economy, it’s simply too new and evolving too fast to know precisely what the impact will be. But I believe there is far more evidence pointing toward hope than demise.
What if AI is actually the catalyst that drives global decarbonization more than any factor before it? Surging power demand is already pulling the next generation of energy technologies to market. It’s responsible for the biggest clean energy power purchase agreements in history, while forcing the bipartisan work of permitting reform. And it’s providing a platform for more rapid innovation in decarbonization technologies throughout the economy.
The opportunity is so clearly there. The actions we take next will help write the optimistic version of this story.
History Tells Us This is a Moment for Clean Energy Innovation
Goldman Sachs Global Institute estimates AI capital expenditures at nearly 8 trillion dollars between now and 2031. That is a far larger undertaking than what was spent on the Manhattan Project, the Apollo program, or the late-1990s telecom boom—each of which led to breakthroughs that still shape our modern lives, including some of the most important clean energy technologies in operation today.
In its race against the Soviet Union, the U.S. space program paid a steep premium for satellite-grade solar photovoltaics that became the foundation of the modern solar industry—now roughly 7% of global electricity generation. The program that produced the nuclear submarine, built to win the same Cold War, became the foundation of the commercial nuclear industry. Today, nuclear is the second largest source of baseload, carbon-free electricity in the U.S. and the world. In each case, a deep-pocketed buyer (the federal government) was willing to pay a premium to achieve their broader objectives, and a new clean energy industry was born in the process. AI is no different. The buildout is driven by the best-capitalized companies in the world—racing against today’s rival, China—and the next-generation clean energy it is pulling toward commercial scale will be the spillover of that competition.
Pulling Demand for All Sources of Electricity
The immediate demand for electrons is driving the build out of everything—renewable, storage, and fossil—creating concerns about near-term rising emissions. In April 2022, the U.S. Energy Information Administration (EIA) projected that the 2025 U.S. power sector would total 3,971 TWh. Actual generation in 2025 totaled 4,522 TWh, about 14% higher than projected. Zero-carbon sources outperformed expectations, but most of the increase came from higher gas dispatch to meet rising demand.
Higher electricity load will also encourage the construction of new, though relatively high-efficiency, natural gas plants and prolong use of a subset of the coal fleet even if most planned retirements are proceeding. In the short-term as demand for power generation surges, data centers are turning to unconventional mobile generators, such as aviation and marine turbines, as they wait to be connected to the grid. Over the next few years, U.S. emissions from power generation will be higher than if AI had never existed in the first place. This is indisputable.
But AI is also the most powerful clean energy demand signal we have ever seen. It is driving the largest renewable energy purchase agreements in history. Amazon has contracted more than 40 gigawatts (“GW”) of clean energy globally. Microsoft’s agreement with Brookfield Renewable will deliver more than 10.5 GW of new wind and solar across the U.S. and Europe by 2030—roughly eight times the size of the largest corporate PPA ever signed at the time. Meta has assembled well over 100 renewable energy projects across its global footprint. Google’s new Minnesota campus will include the largest battery project by capacity in the world, along with huge wind and solar builds, as part of a clean energy portfolio that’s at a comparable scale to its peers. Indeed, the hyperscaler clean energy portfolio represents the most concentrated corporate energy procurement effort in history. The four tech companies account for a dominant share of the nearly 100 GW in clean energy capacity the private sector has contracted in the U.S. since 2022.
This unprecedented procurement of clean energy is impacting forecasted national energy capacity additions. Even under the EIA’s most recent high electricity demand scenario, roughly 90% of net capacity additions through 2035 are projected to come from zero-carbon resources (including storage).
Extend Use of Existing Clean Energy Resources
AI’s electricity demand is also giving extended life to the most important zero-carbon baseload asset we already have: nuclear. Microsoft signed a 20-year agreement to restart Constellation’s Crane Clean Energy Center. Google followed with a 25-year agreement to restart NextEra’s Duane Arnold Energy Center in Iowa, a power plant that closed in 2020 and is expected to be back online by early 2029. Meta signed 20-year PPAs with Vistra for three nuclear plants in Pennsylvania and Ohio that were on a path to retirement as recently as 2020. Baseload, zero-carbon power is being given a new life by the force of AI demand. The last time the U.S. let nuclear capacity stagnate, the zero-carbon power it would have provided was replaced overwhelmingly by coal for the next 15 years.
There’s also a major opportunity to deploy technologies and buyer-seller arrangements that squeeze more efficiency out of the grid we already have. Demand response programs (paying large consumers to power down when demand peaks), reconductoring (replacing existing lines with higher-capacity conductors), and battery storage remain underutilized, though deployment is accelerating. Better forecasting of supply and demand, weather, and line capacity, plus a host of other software tools—all improved by AI—will help us get significantly more out of the system we have.
Competing for the Next Generation of Energy and Economic Leadership
Another hard truth: data centers are going to be built. The only question is where. China has made AI dominance a national priority, and its AI will be powered prominently by coal. India and Malaysia are emerging data center hubs whose grids are also heavily dependent on coal. The Climate Leadership Council has spent considerable time studying the relative emissions of grids around the world, and I am confident that in most cases, a data center built outside the United States will produce more emissions than one built here. If the U.S. fails to build the energy infrastructure to power AI domestically, data centers—and the technological breakthroughs they can generate—will move to grids where the carbon math is worse.
I know this argument does not move many in the U.S. climate community. “Better-than-China” is rarely a winning rallying cry for domestic environmental activists. The facts are still true, but if you’re not compelled, fair enough.
There’s another important dimension to the competitiveness conversation, too. The U.S. has a long history of producing world-changing ideas. Unfortunately, it has not always been the primary benefactor of its ideas. Remember how the U.S. space program essentially created the modern solar industry? Solar is now dominated by China, which controls more than 80% of global share in the manufacturing process. If we shy away from this moment and the clean energy technology development that will come with it, the U.S. risks ceding its competitive standing yet again. Future energy dominance requires being the leader in these next generation technologies, as does our future energy security—not to mention jobs and affordability.
Demand will Deliver Permitting Reform
Larger debate aside, most people in policy agree that the single biggest obstacle to building more clean energy is permitting. The current system for approving how new power supply will move from source to destination is outdated, cumbersome, and massively time intensive. Analysis has found that decarbonizing the power sector would require the U.S. to more than double its regional transmission capacity and expand interregional transmission by more than fivefold.
Even without the advent of AI, powering our nation in an increasingly electrified economy demands that utilities, grid operators, and policymakers confront problems that have long been deferred. Permitting reform is coming, and AI is becoming the forcing mechanism to finally create enough political will for action.
A Moment that Could Change Everything
The biggest reason for optimism that AI is a catalyst, not a climate catastrophe, is the innovation it will drive.
Global decarbonization requires the development and commercialization of breakthrough technologies capable of providing clean, firm power like fusion, long-duration storage, advanced geothermal, next-generation nuclear, and carbon capture (among the most promising). The path from prototype to commercial scale gets dramatically shorter when there is a large, creditworthy buyer willing to sign a long-term contract.
AI is becoming that buyer. Microsoft signed the first-ever fusion PPA with Helion Energy, which was followed by an agreement by steelmaker Nucor with the fusion company. Google signed the first-ever corporate agreement for advanced geothermal with Fervo Energy (expanded in 2024), and a first-of-its-kind PPA for a 400 MW gas plant in Illinois paired with carbon capture to sequester roughly 90% of the plant’s CO₂ underground. For decades, the case for building CCS stalled without a customer. AI is providing one. Amazon partnered with X-energy to develop small modular reactors in Washington, part of its 5 GW nuclear power plan over the next 15 years. Meta committed to up to 1.2 GW of Oklo advanced reactors in Ohio and additional capacity from TerraPower’s Natrium reactors. These are not research grants. They are binding commercial commitments or direct investments pulling next-generation clean energy toward viability.
AI is not just the customer for these technologies. It is also a tool to develop them faster. Four tech majors just teamed upwith Elemental Impact to use data centers to accelerate the development of technologies like low carbon cement and steel, energy storage, and advanced cooling. Researchers at MIT and the national laboratories have identified energy storage, fusion plasma control, and carbon capture as areas where AI could deliver step-change improvements. We are on the verge of bringing the power of the sun to the Earth, and AI is going to help us get there sooner.
The Work Ahead
A once-in-a-generation demand signal, breakthrough innovation, and transformative volumes of capital investment are rapidly pulling next-generation clean energy toward commercial viability. I believe we will look back at this moment, a decade or two from now, and say this was when game-changing technologies accelerated to the finish line.
Policy will matter. Corporate leadership will be essential. The incentives have to be right—and getting them right is the work of the Climate Leadership Council. This moment calls for bringing together stakeholders from across the economy, members of Congress on both sides of the aisle, and decision-makers throughout government to confront the hard questions and find solutions. Solutions that unleash the power of the market, provide a level playing field, remove barriers to building the infrastructure we need, and ensure the U.S. is the global leader in the next generation of clean energy technologies.
There are real challenges ahead, and hard conversations to come. You have to go through it to get to it. The path to global decarbonization runs through this moment, not around it.