AI data centers face energy and cooling supply crunch

The AI infrastructure narrative has evolved beyond semiconductor supply to focus on the power, cooling, and energy-storage infrastructure required to run massive AI data centers. Softbank's recent multi-billion-dollar investment in AI data center batteries has crystallized investor focus on the next supply constraint: reliable, clean power and advanced cooling systems. Data centers are energy-intensive and require both stable AC power and liquid or immersion cooling to manage heat loads that rival small power plants. This realization is driving investor interest in battery storage names like BE (Bloom Energy), EOSE (Eos Energy Enterprises), and FLNC (Fluence Energy Integrator), as well as HVAC and thermal management plays.

Target Hospitality announced a multi-year contract expected to generate over $750 million in revenue to develop an "AI Infrastructure Community" with vertically integrated power and cooling infrastructure. Fervo Energy, a geothermal developer backed by Bill Gates, has raised its IPOInitial Public Offering - a company's first public sale of stock. target to $1.8 billion, positioning geothermal as a key renewable-energy solution for 24/7 baseload power to data centers. VoltaGrid, a power-management startup, just closed a $1 billion Series D co-led by Blackstone and Halliburton, valuing the firm at more than $10 billion. Duke Energy applied for Department of Energy loans "representing potentially billions of dollars in customer savings," signaling that traditional utilities are pivoting toward AI-era infrastructure investment.

The thesis is straightforward: AI chip buildout is only one piece of the puzzle. Data center operators face power availability constraints in key regions like Northern California, Texas, and the Midwest. Energy costs can represent 20-30% of data center operating budgets. Cooling technology must evolve to handle liquid immersion and advanced thermal management. Companies selling the "picks and shovels" for data center power and cooling are poised to benefit more durably than commodity chipmakers facing cyclical overcapacity.

Skeptics argue that energy and cooling represent mature, capital-intensive industries with thin margins. If hyperscalers invest directly in power generation and cooling systems rather than outsourcing, stand-alone energy and cooling vendors could see margin compression. Additionally, if AI capex disappoints or consolidates to a narrower set of mega-cap hyperscalers, demand for specialized energy and cooling infrastructure could plateau.