Training large language models consumes roughly 15 megawatts per data center cluster, while inference at scale adds continuous baseload demand that grid operators cannot reliably meet. Cooling alone accounts for 40 percent of data center energy budgets. This convergence—peak demand from hyperscalers colliding with aging grid infrastructure and renewable intermittency—has created an acute supply constraint. Antora Energy is betting it can solve this by deploying thermal energy storage systems that capture waste heat from data centers and convert it back into electricity during peak demand windows. The $550 million Series C, one of the year's largest cleantech rounds, signals that venture capital now views energy infrastructure as a critical unlock for AI's next growth phase rather than a peripheral concern.
Antora's thermal storage technology differs fundamentally from lithium-ion batteries. The startup uses molten salt systems to store heat at high temperatures, then converts that stored energy back to electricity when needed. This approach offers two advantages over conventional batteries: lower per-megawatt-hour costs at grid scale and compatibility with existing data center thermal loops. Competitors pursuing adjacent infrastructure plays include Form Energy, which raised $240 million Series D for long-duration iron-air batteries, and energy dispatch software companies like Stem and Sunrun, which optimize grid-scale storage deployment. Each addresses different nodes in the same problem: how to match explosive AI compute demand with reliable, affordable power.
The pattern mirrors the late-1990s fiber optic buildout, when venture capital mobilized $150+ billion into infrastructure after recognizing that bandwidth constraints would strangle internet adoption. Within three years, that conviction flipped: software VCs fled traditional infrastructure and followed capital downstream to application layers. Today's dynamic is inverting. Traditional energy VCs are co-investing alongside growth-stage software funds—Antora's round included backing from both greentech specialists and AI-focused venture firms—suggesting that software-native investors now view infrastructure as non-negotiable to their AI exposure. This reshaping has immediate downstream consequences: generalist VCs without infrastructure expertise risk mispricing Series A and B rounds in power, cooling, and grid software, while infrastructure specialists are graduating from energy-focused funds into top-tier AI portfolios. For founders, it means capital availability for the unsexy but essential layers beneath model training is higher than ever.