Teravolt looks to cannibalize older industries to meet AI power demand
Bitcoin farms, aluminum smelters, and other old infra is more lucrative to repurpose as a datacenter
Teravolt, an AI infrastructure company based in London, seeks to adapt older industries to meet the growing power demands of the AI sector. The current AI market is facing challenges, with US hyperscalers incurring $220 billion in debt and major AI model creators yet to demonstrate profitability. Public concern over the development of datacenters required to support AI demand adds to the uncertainty.
According to Gartner, datacenter power demand will increase from 104 GW in 2025 to 132 GW in 2026 and reach 290 GW by 2030. Morgan Stanley predicts a potential shortfall of 49 GW by 2028, with the need for 410 GW by 2036. Building new grid infrastructure is a lengthy process, taking five to 15 years for planning, permitting, and construction, which makes it difficult to meet the rapid AI power demand.
To bridge this gap, Teravolt proposes repurposing existing energy assets, such as old power plants, industrial sites, or refineries. These assets already possess infrastructure, permits, contracts, and staff, making retrofitting an attractive option. Laert Karaashev, co-founder and managing partner of Teravolt, expects AI components to become commoditized, except for the power aspect, which is challenging to commodify.
As AI workloads become more valuable than other industries, the company believes that some AI projects may not materialize, workloads may shift to areas with surplus energy, and some power generation infrastructure will remain longer than planned.
Written by urgent.news from The Register Science's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
