AI data centers face power and cooling crunch

As AI capex accelerates, the hard constraints of physical infrastructure are becoming apparent. Every megawatt of hyperscale data center capacity requires roughly 27 tonnes of copper for transformers, substations, power distribution, cooling systems, and high-capacity cabling. This insatiable appetite for copper, combined with China's sulphuric acid export restrictions, has created a crunch. Metals.co is now exploring deep-sea nodule mining to scale copper production, and mining companies are pivoting toward acquisitions in frontier markets.

Cooling is an equally acute bottleneck. Data centers consume massive amounts of water and require cutting-edge thermal management tech. Battery and energy-storage firms like Altus Energy, Battery Energy Storage, and others are seeing demand surge as operators seek to buffer grid volatility and store renewable power. SoftBank's recent billion-dollar investment in AI data center batteries signals institutional recognition of this constraint. Investors are now hunting for companies with innovative cooling and battery-storage solutions.

Power procurement remains the ultimate chokepoint. Grid operators in Europe and the US are warning that local transmission capacity is inadequate for the projected AI buildout. Some data-center operators are eyeing nuclear power deals; others are pursuing geothermal (Fervo Energy's IPOInitial Public Offering - a company's first public sale of stock. target raised to $1.8 billion) or renewable contracts with a 20-30 year horizon. The transition from capex euphoria to infrastructure reality is beginning to reset expectations for AI margin profiles and timelines.

Investors are rotating from pure AI compute names into picks on the supporting infrastructure: copper miners, cooling-tech firms, battery-storage companies, and grid-reinforcement plays. The narrative is shifting from "unlimited AI buildout" to "AI buildout constrained by physics and permitting."