as artificial intelligence With rapid expansion and soaring demand for energy and investment, the market may struggle to keep up. If providers cannot supply and finance enough energy to power AI, and policymakers cannot put in place regulations to reduce broader financial risks, entire sectors and downstream economies could collapse.
This warning was the focus of a panel discussion held on February 5th at the Institute of Politics at the Harvard Kennedy School (HKS). There, policymakers and economists examined the scale and risk profile of AI investments, as well as their impact on the U.S. energy sector, climate policymaking, and the environment itself.
Joseph Aldi, Heinz Professor of Environmental Policy at HKS and former White House climate advisor, opened the discussion by outlining the growing environmental costs of AI. Demand for electricity in the United States is rapidly increasing, in part due to the rapid proliferation of data centers. At the same time, political resistance is slowing the construction of technologies that can ease the burden, such as wind and solar power.
Mr Aldi said the result was a troubling contradiction: “We will not be able to build enough gas-fired power plants to make up the difference.” Coal-fired power plants once expected to be retired are still operating, power prices are rising, and emissions are beginning to rise.
Aldy pointed to the potential of AI to accelerate the discovery of energy-efficient materials, improve climate modeling, optimize the power grid, and even support carbon removal. But he cautioned that these environmental benefits are not immediate or guaranteed. And without a clear policy signal to steer AI into climate-related applications, Aldi said investment will likely flow to what has the quickest returns, which may not be green energy.
Robert Rubin ’60, who served as Treasury secretary under President Bill Clinton and the first chairman of the National Economic Council, placed the AI infrastructure investment boom within a broader macroeconomic cycle that is itself increasingly driven by AI. He warned of what he called “cyclical risk,” in which large AI companies make large future commitments to suppliers, prompting suppliers to borrow aggressively, invest in production capacity, and return capital to the same companies where demand forecasts justify expansion.
Rubin pointed to OpenAI, which has approximately $1.4 trillion in long-term contracts with suppliers. “The supplier’s stock price will go up because of the promise,” he explained, explaining that higher valuations will provide more funding to the AI companies themselves. OpenAI now faces rapidly increasing competition from other companies and platforms (such as Anthropic, DeepMind, Meta, and Google), making it increasingly likely that it will not be able to deliver on its promises, spelling disaster for its stakeholders, the sector as a whole, and the many other industries it currently serves.
Jason Furman, an Aetna professor specializing in economic policy practice at HKS, suggested adding an extra layer to the simple boom-or-bust interpretation. Furman said investment returns will still increase significantly, but may be lower than today’s most optimistic forecasts.
Furman pointed to a study by Federal Reserve economists that tracks real estate listings and early construction activity to estimate future data center capacity one to two years out. “There’s a tremendous amount of data on what started,” Fuhrman said. He said these indicators suggest that AI-related investments are likely to continue to grow, although the growth rate may not match the most alarming predictions from figures like Elon Musk and Sam Altman.
Aldi added that much of the available data is inconclusive, partly due to a “broken regulatory system”. Long and inefficient approval processes lead companies to hedge their risks by submitting multiple proposals at once, assuming only one will likely clear the regulatory hurdles.
In regional power markets like PJM, which serves much of the Mid-Atlantic region, the rapid growth of data centers is contributing to soaring prices, reigniting the debate over whether consumers or businesses should pay for new infrastructure. Aldi pointed out that the core problem is that AI companies are large power users and don’t face the full cost of imposing on their systems.
Towards the end of the session, a first-year public policy master’s student asked how a government like the United States would make policy in a world 30 to 50 years from now, in a world that is largely “agent” (i.e., AI agent-driven). In an economy dominated by autonomous systems, she asked, what are the incentives for people to participate in the first place? Will traditional economic structures be replaced by “more relational ones”?
As Rubin puts it, the challenges ahead hold “tremendous potential, both positive and negative.” Whether the United States obtains the former or succumbs to the latter may depend more on political will than on innovation, he suggested.
