AI Data Center Demand Strains Electricity Markets

The AI boom is hitting real infrastructure limits. US power prices climbed 61% faster than inflationThe rate at which prices rise across an economy. as demand surges, with consumer electricity climbing last month by the most in three years. American Electric Power, one of the biggest US utilities, is seeking to raise $2.6 billion in a share sale as artificial intelligence technology drives booming demand for power. Alphabet is making its first-ever bond sale in yen, broadening funding channels as the Google parent sharply increases capital spending to finance artificial intelligence infrastructure.

Hardware and semiconductor makers are racing to keep up. NVIDIA, AMD, and Broadcom have dominated retail trading interest, with call-to-put ratios at extreme levels. Bitcoin's hash rate just dropped 4% in its first negative growth quarter in 5+ years, a sign that AI compute competition is crowding out mining operations. Meanwhile, semiconductor stocks are near 52-week highs despite a modest pullback, with the Semiconductor Index up nearly 73% year-to-date.

The infrastructure bind extends to data center real estate and power procurement. Private equity is becoming more selective on real estate bets, with Rockpoint signaling capital is flowing selectively to compute-intensive assets. Utilities, grid operators, and equipment makers are all racing to scale, but the reality is clear: supply-chain constraints in power delivery and chip fabrication will persist through 2026. Energy generators and transmission firms benefit from elevated capex; equipment makers and utilities see margin expansion near-term before competition pressures pricing.

The debate centres on sustainability. Some market participants worry that the power crunch will force AI companies to shift capex away from hyperscalers toward smaller, distributed infrastructure. Others note that electricity supply can adapt over time through renewable energy additions and grid modernisation. The real risk is if capex disappointment hits once the market realises power limitations constrain AI deployment velocity.