The artificial-intelligence boom is entering a more difficult phase: its biggest constraint is no longer only chips or capital, but electricity. Across the United States and other major economies, the rapid construction of data centres is driving up demand for power, water and transmission capacity. That pressure is now becoming a political liability, while investors are questioning whether enormous spending on AI infrastructure will generate returns quickly enough.
The tension was exposed after Democrats won sweeping victories in the U.S. midterm elections. Market commentary linked the subsequent technology sell-off to fears that the new political majority could slow data-centre construction, particularly in states where opposition to large facilities had become a campaign issue. Reuters reported that at least $3 trillion in S&P 500 market value was wiped out during the tech-driven rout, although the decline also reflected broader concerns about high valuations, debt and the cost of the AI buildout.[Reuters]
Why the boom is running into resistance
Data centres are capital-intensive facilities that require reliable electricity around the clock. Their local effects can be immediate: new substations and transmission lines, competition for land and water, construction disruption and potentially higher costs for consumers if utilities expand infrastructure on the assumption that projected demand will materialize.
Supporters of rapid development argue that AI can raise productivity, attract investment and strengthen national competitiveness. Technology companies and their suppliers have therefore committed vast sums to servers, networking equipment and power contracts. Meta, for example, said it would raise another $15 billion through a bond offering to finance AI expansion, while Marvell announced a $115 billion all-stock acquisition of optical-components maker Lumentum, including debt.[Reuters]
Critics do not necessarily oppose AI itself. Their argument is that private firms should not receive an implicit public subsidy through socialized grid costs or relaxed environmental rules. Local moratoriums and permitting restrictions reflect concerns that communities are being asked to absorb the infrastructure burden before the economic benefits are clear.
The investor question
The market anxiety goes beyond regulation. AI companies are spending at a scale that assumes continued demand for computing, rising revenues and sustained technological leadership. That assumption is increasingly being tested. Reuters reported that investors were already concerned about whether massive AI investments would produce sufficient returns before the U.S. election intensified worries about data-centre policy.[Reuters]
A more cautious market does not prove that AI is a bubble. It does show that investors are beginning to distinguish between useful applications and infrastructure spending justified mainly by expectations of future growth. Higher borrowing costs make that distinction sharper, especially as governments carry heavy debt and companies raise money in increasingly volatile markets.
“The technology's rapid rollout is driving a surge in global power demand, contributing to rising greenhouse gas emissions.”[Reuters]
Climate policy adds another layer
The conflict is particularly visible at COP31 in Antalya, Turkey, where negotiators are seeking progress on electrification and a renewed shift away from fossil fuels. Turkey’s presidency is promoting a target for electricity to supply more than one-third of global energy use by 2035, compared with roughly one-fifth today.[Reuters]
Electrification can support decarbonization if power systems add sufficient renewable, nuclear or other low-carbon generation. But AI demand complicates that transition. War-related energy insecurity has already encouraged some countries to secure oil cargoes and restart coal plants, while data-centre growth adds a new source of electricity consumption. The result is a race between clean-power investment and demand growth.
Business leaders have urged COP31 participants to endorse faster electrification and a 35% target for 2035. Their projections suggest that without stronger policy support, global warming could reach 2.6°C by 2100; expanded clean electrification and a phase-down of coal could lower that estimate to about 2.3°C.[Business leaders’ COP31 letter]
What comes next
The next phase will likely be defined by bargaining rather than a simple victory for either technology companies or environmental campaigners. Governments may require developers to finance grid upgrades, secure cleaner power and disclose water use. Utilities will face pressure to prevent new industrial demand from shifting costs onto households. Companies, meanwhile, may prioritize fewer, more efficient facilities and seek stronger evidence that AI workloads can support their valuations.
The political lesson is equally clear: voters may welcome technological innovation while rejecting its local costs. If AI is to retain public support, its gains will need to appear in jobs, services and productivity—not only in technology-sector share prices. The industry’s challenge is to prove that the infrastructure being built today is an economic foundation rather than an expensive bet on tomorrow.
Sources: Reuters on COP31; Reuters on U.S. markets and data centres; Reuters on AI investment risks; COP31 business letter.