Teravolt looks to cannibalize older industries to meet AI power demand

If you are building or deploying AI infrastructure, you may face higher energy costs or longer lead times due to repurposed industrial power sources.

Teravolt, an AI infrastructure company, proposes repurposing existing energy-intensive industries like Bitcoin farms and aluminum smelters to meet the immense power demands of AI computation. This strategy addresses the significant projected shortfall in electricity supply needed for AI growth. The company cites Gartner predictions of global data center power demand reaching 290 GW by 2030 and Morgan Stanley's estimate of a 49 GW shortfall in the US by 2028. Teravolt notes that building new grid infrastructure takes significantly longer than repurposing existing industrial sites, which already possess some level of infrastructure, permits, and staff. The company's co-founder, Laert Karaashev, suggests that power is the hardest part of the AI stack to commoditize, making the cannibalization of older industries a faster route to meet demand, especially as AI compute generates substantially higher revenue than traditional industrial output. Teravolt is currently focusing on brownfield sites in Eastern and Southern Europe due to their lower costs and faster development timelines compared to Western markets, which are necessary to meet customer demands for GPU deployment within 12 months.

Teravolt's strategy highlights the burgeoning crisis in energy supply for AI, suggesting that the demand for power is rapidly outstripping the capacity of traditional grid expansion. By targeting industries with established power infrastructure, Teravolt aims to bypass the lengthy permitting and construction timelines associated with new grid development, effectively turning an energy deficit into an opportunity for existing power assets. This approach underscores the commoditization of AI compute itself, with power emerging as the critical bottleneck, driving innovative, albeit disruptive, solutions.

The company's focus on Eastern and Southern Europe indicates a strategic pivot driven by economic feasibility and customer urgency, as Western markets present significantly higher costs and longer delays. This geographical shift signals a potential reshaping of AI infrastructure deployment, favoring regions with readily available brownfield sites and less regulatory friction. As AI workloads become more valuable than traditional industrial output, we can expect increased competition for these repurposed energy assets, potentially driving up acquisition costs and further concentrating AI development in specific, strategically advantageous locations.