AI Capex Surges, But Power Prices And Governance Rules Are Starting To Bite

The AI boom has entered a more expensive and more disciplined phase in 2026. Reuters reported on 31 March that Big Tech’s AI spending plans are facing an energy shock test, with a combined 635 billion dollars of infrastructure and cloud spending now colliding with higher power costs and tighter investor scrutiny.​

That is a huge amount of capital, and it is changing the technology landscape from the ground up. Data centers need power, cooling, land, networking and chips, and every one of those inputs is being affected by the current energy shock. The bigger the build-out, the more sensitive returns become to electricity prices and project delays.

The market is already adjusting. Reuters has described bond-market pressure on Big Tech as investors increasingly question whether the AI spending cycle can sustain valuations. This is happening at the same time that software and data firms are coming under pressure from AI commoditization and the risk that customers will build more of their own tooling instead of buying premium software.

Governance has become the other major issue. Reuters has warned that AI accountability is still lagging, and Asian regulators are now enforcing stricter laws in markets such as Vietnam and South Korea. That means technology companies can no longer treat AI rollouts as purely a product decision; they must build compliance, labeling, human oversight and risk management into the product from day one.

The long-term outlook is still bullish, but the winners are getting narrower. Semiconductor equipment, grid infrastructure, power management, cooling and enterprise-grade AI solutions are all likely to benefit more than generic software plays. The era of “AI at any price” is fading, and the market is beginning to insist on proof of economic value.