Artificial intelligence firms are exploring dedicated power supplies for data centers as mounting political pressure over rising household electricity bills threatens to slow the rapid expansion of AI infrastructure, Wood Mackenzie strategists said in a note on Monday.
Ed Crooks, vice chair Americas at Wood Mackenzie, said that major technology firms are exploring alternative power supplies that can operate independently of the grid as opposition to new data-center developments gains traction across the US.
Lawmakers in 14 US states have debated restrictions on new data center developments, reflecting growing bipartisan concerns over the industry's energy consumption.
Bans are being considered in Democratic states such as Vermont and New York, in Republican states such as Oklahoma and South Dakota, and swing states such as Pennsylvania and Georgia, said Crooks.
Though electricity accounts for only a small share of the total cost of AI services, concerns about household affordability have emerged as a significant political challenge for the sector.
Wood Mackenzie said political pressure over household electricity bills is adding to the pressure to develop collocated power supplies.
The consultancy said that major technology firms have pledged to cover the costs of new generation capacity and grid upgrades needed to support their operations.
However, regulatory frameworks in most US power markets are designed to spread system costs across all consumers, making it difficult to isolate the impact of individual data centers on electricity prices.
Wood Mackenzie analysts Ben Hertz-Shargel and Chris Seiple said data centers connected to the grid generally benefit from lower costs and greater reliability than facilities relying on dedicated generation assets.
However, lengthy delays in obtaining grid connections are prompting some operators to pursue "bring your own power" strategies, in which generation is built alongside data centers.
The data center companies that can operate reliably without firm grid service will be able to scale their AI businesses faster than others, the analysts said, adding that such flexibility could provide a competitive advantage in the race to deploy AI infrastructure.
Meanwhile, Wood Mackenzie said that hyperscale cloud providers are increasing efforts to support energy innovation.
Microsoft, Amazon, Google, and Meta recently announced a joint initiative with non-profit investment group Elemental Impact to fund emerging energy and materials technologies that could reduce the environmental impact of data centers.
The program aims to support up to 10 technology startups by the end of 2027. However, the consultancy said that new energy technologies are unlikely to be deployed at a sufficient scale to materially affect AI-related electricity demand within the next decade.
Wood Mackenzie said even advanced technologies such as small modular nuclear reactors and fusion power are projected to play a limited role in the US power mix before 2040.
AI firms seeking rapid expansion are expected to rely on existing technologies and infrastructure while balancing growing scrutiny from regulators, consumers and environmental groups over energy use.