The AI infrastructure boom is running into a problem that has little to do with demand. Political resistance to new data centers is growing as communities worry about electricity grids, water consumption and environmental pressure. Investors are now asking whether slower construction could cool one of the market’s biggest growth stories.
Data centers hit resistance
Jim Cramer recently described the AI trade as “broken,” pointing to growing opposition to data center development. Computing demand remains strong, but securing permits, power connections and community approval could become harder. That creates uncertainty for suppliers such as GE Vernova, which has benefited from soaring demand for power infrastructure and now has a $76 billion backlog.
Construction delays, however, could push projects further out and slow expected revenue growth. Semiconductor companies face similar pressure. Nvidia remains dominant in AI accelerators, while AMD is preparing its MI400 GPU roadmap and Micron benefits from demand for high-bandwidth memory. Recent volatility suggests investors are becoming more selective, with AMD shares falling 4% during a broader semiconductor pullback.
Power becomes crucial
Electricity infrastructure could prove more resilient than some parts of the AI trade because data centers require enormous amounts of power. GE Vernova, Eaton and Cummins have all benefited from equipment demand linked to the expansion. Higher Treasury yields add another complication. Growth stocks typically depend heavily on expectations of future earnings, making their valuations more sensitive when discount rates rise.
Longer term, the underlying appetite for AI computing has not disappeared. Wider adoption of large language models, computer vision and autonomous systems continues to support infrastructure demand. Still, the easy phase of the AI investment boom may be ending. Political resistance, power constraints and higher yields could make 2026 less about buying anything connected to AI and more about identifying companies capable of navigating a slower, more complicated buildout.
