A race is underway among the world’s largest corporations for dominance in artificial intelligence and, so far, investors are along for the ride. Leading artificial intelligence (AI) chip maker NVIDIA
SPDR Dow Jones Industrial Average ETF Trust
This investment frenzy rests on the thesis that AI will soon unleash productivity gains across the economy. With its enormous data processing capabilities, AI will inevitably transform life and work in our information-driven economy. Because we are just beginning to comprehend its benefits and risks, a cautious approach with significant industry oversight is warranted.
AI proponents highlight the promise of new grid management and other decarbonization tools. However, at the moment, the AI industry’s most significant impact on the climate has been to unleash new fossil fuel power generation and associated increases in greenhouse gas emissions. This worsens the risk of a climate change-spurred economic crash that would dash the AI industry’s hopes of building a strong customer base and achieving profitability.
Action is needed by tech and utility industry leaders, investors, regulators, and policy makers to ensure that AI’s buildout is powered by clean energy sources. This will not only limit climate-related damage to the global economy, it will help the AI industry help itself.
Data Center Construction Boom Fueled by AI
In a 2023 analysis of U.S. grid planners’ projections, energy analysts Grid Strategies reached a startling conclusion: five-year projections of electricity demand growth had increased from 2.6% in the previous year to 4.7%.
Much of this projected growth is connected to new data centers. McKinsey estimates that data centers will comprise 30-40% of all new net demand until 2030. Moreover, data center growth is strongly linked to AI. Goldman Sachs estimates that by 2028, AI will represent about 19% of all data center power demand.
Other important drivers of the U.S. electricity demand increase are industrial reshoring and electrification of transport, buildings and industry, both central planks of the worker-friendly energy transition agenda of the Biden-Harris administration, incentivized by the landmark Inflation Reduction Act.
Among these energy demand drivers, AI is unique in that a competitive advantage tends to go to those companies that use more energy (by processing more data, AI companies can train larger and potentially more capable models). This suggests that the rapid acceleration of AI’s energy consumption is not likely to end anytime soon.
AI’s Unmanaged Climate Risks
Forecasts of rapid AI-fueled energy demand growth have prompted utilities to delay planned closures of coal-fired power plants. S&P Global recently projected that the U.S. power sector would take offline just 105,000 megawatts of coal-fired power by 2035, a 21% decline in retirements projected a year earlier.
Gas-fired power plants, once threatened by the rise of inexpensive renewable power and batteries, also have received a new lease on life. More new gas power capacity was announced in the US in the first half of 2024 than in all of 2020.
Unless a new path is chosen, this resurgence of fossil fuel power could represent a major setback to the energy transition and efforts to prevent climate breakdown.
Avoiding New Fossil Fuel Infrastructure This Decade
Tech industry leaders are pledging to address these climate risks, but mostly with investments that will bear fruit in the 2030s and beyond, outside the time period in which scientists say aggressive emissions reductions are most essential. For example, last month Microsoft pledged to bring online power from a restarted Three Mile Island nuclear facility owned by Constellation Energy, and this week Google announced plans to purchase power from small modular reactors to be developed by the startup Kairos Power.
Inflation Reduction Act (IRA) production tax credits are key to making these projects financially viable. Assuming nuclear waste challenges are addressed, they could represent important IRA successes in driving the decarbonization of the U.S. electricity system. But they would do nothing to prevent climate damage from new fossil fuel infrastructure built this decade. Fires, floods, and other disasters linked to greenhouse gas emissions are already escalating rapidly, and production from existing fossil fuel projects already exceeds Paris-aligned consumption levels.
The tech and utility industries must together aggressively curtail the gas plant buildout, proceed with scheduled coal plant retirements, and accelerate efforts to deploy renewable energy and energy storage. This would make a meaningful dent in systemic climate risks and help the U.S. meet its Paris Agreement commitments.
Investors Must Grapple With Climate Science To Fulfill Their Fiduciary Duties
Heading off the AI-fueled spike in emissions requires countering the myth that climate change is solely the responsibility of governments. Institutional investors have a fiduciary duty to ensure that companies address systemic climate-related financial risks. This is especially true of the many investors with diversified portfolios that are heavily exposed to economy-wide climate shocks.
Climate change particularly jeopardizes AI profitability. Unlike other new technologies, to become profitable AI must bring in revenues sufficient to offset large annual capital expenditures. Key AI industry observers, such as Jim Covello, Goldman Sachs’ stock research head, question whether, with the reliability challenges that limit its usefulness, AI will ever secure these revenues.
The AI investment case in response is that companies in virtually every industry sector will want to find out how AI might provide them with a competitive edge. But if a significant downturn in the economy emerges, this widespread adoption becomes much less likely. Many potential customers may not be willing or able to add significant expenditures to their budgets for an unproven technology that struggles with reliability.
Addressing the risks of a potential climate change-induced crash of the economy on AI’s customer base is now a business imperative for the nation’s largest companies and their investors.
How much attention to the risks of a climate-related crash is warranted? At a minimum, investors must be attuned to emerging climate science and its implications. According to an October 2024 analysis from the EDHEC Risk Climate Impact Institute, global equity valuations could drop as much as 40% if emissions reductions do not accelerate. This projection does not factor in tipping points, such as the collapse of the West Antarctic glacier and the thawing of methane-rich Arctic permafrost, that if reached would further destabilize the economy.
Proposals For An AI-Focused Carbon Price
Tech industry analyst Robert Wright argues that OpenAI CEO Sam Altman has put us on a dangerous course by promoting “the idea that, when it comes to technological change, and progress in AI in particular, faster is better.” He proposes that governments refrain from subsidizing AI buildout until its climate and safety risks are addressed, and instead impose a special tax on power consumed by AI data centers to help temper these risks.
Similarly, the Biden-Harris Administration’s former National Economic Council director Brian Deese proposes new policy incentives for utilities and tech companies to bring online clean energy to power AI’s massive computing needs. A key recommendation is that the federal government impose a fee on the tech industry to be paid into a clean energy deployment fund at the Department of Energy.
Any hope for Congressional action on these ideas would require tech industry leadership on carbon pricing that to date has been missing.
Near-Term Actions to Decarbonize The Electricity System
Fortunately, meaningful actions to decarbonize the electricity system outside of Congress are feasible. Opportunities for tech and utility industry leadership include expanding procurements of around-the-clock carbon-free energy, increasing transparency about projected demand, enacting green energy tariffs, and supporting interregional transmission and other grid modernization.
Procuring around-the-clock carbon-free energy is perhaps the most contentious of these actions because it is more expensive than the traditional approaches of buying renewable energy wherever it is cheapest. But as shown in a 2024 analysis by Princeton University’s Jesse Jenkins and colleagues, traditional procurement approaches have not proved effective in reducing emissions. Only by procuring their clean energy at the same time as they consume energy will large customers drive investments beyond wind and solar to include energy storage and firm clean energy technologies such as geothermal and hydropower.
Google has shown important leadership in the movement toward around-the-clock carbon-free energy procurement. Its advocacy for carbon-free energy generation and transmission in the Santee Cooper Integrated Resources Plan is an example of how tech companies can proactively advocate for near-term emissions reductions, not just those in future decades.
Fairness to Ratepayers, Taxpayers, and Communities
Data centers are driving up the market price of energy and will likely continue doing so until interconnection queues and other energy supply constraints are resolved. Tech companies, utilities, and state regulators have a responsibility to ensure the costs of new power plants, transmission, and other system upgrades necessitated by the AI-fueled data center boom are borne by the tech companies that profit from the data centers, not other ratepayers.
Attempts to force ratepayers to cover the costs of the system upgrades for data centers are already underway. For example, utilities AEP and Exelon recently filed an objection to a nuclear power proposal by Amazon on the grounds that it could shift up to $140 million in transmission costs to everyday ratepayers. Similarly, Maryland’s Office of People’s Counsel recently objected to a proposal by the regional grid operator PJM to impose on Maryland ratepayers a substantial portion of a $5 billion capital expenditure to build transmission for new data centers in Virginia.
Regulators and policy makers must also ensure that tech companies avoid and minimize the local environmental impacts of the AI-fueled data center boom. Fossil fuel companies have a long track record of dodging accountability for their environmental impacts, especially in the low-income communities and communities of color that have long served as energy production sites. At least 120,000 oil and gas wells across the US have been abandoned by their original owners and lack financially solvent operators. Community residents suffer the health impacts of this toxic legacy and taxpayers ultimately bear the cleanup costs. Regulators and policy makers must learn from this history and ensure that the tech and utility industries take full responsibility for the environmental impacts of AI’s energy infrastructure.