Fuse is buying gas engines and co-locating AI compute to cut gas power costs about 25%
What's the thesis? On 9 November 2026, Fuse Energy CEO Alan Chang argued that renewables have made gas power more expensive per megawatt-hour, not cheaper. A gas engine that only runs about 20% of the year still carries the same fixed costs, so the cost of each gas-powered MWh rises. Fuse started in grid-scale solar (about 100 MW live and under construction). Because solar only works when the sun shines, Chang says the company is now buying operational gas engines and building new ones too.
The utilisation fix. Fuse's answer is to put Fuse Compute, its AI compute business, next to the engines. Power cost is a small fraction of GPU revenue, he writes, while power availability is the bottleneck. Co-locating AI compute lets the engines run at nearly 100% utilisation instead of 20%, which Chang says cuts the cost of every gas-powered MWh by about 25% and feeds through as lower costs for Fuse energy customers.
Context. Fuse already markets Fuse Compute as single-tenant Nvidia GPUs in UK data centres, with a 142 MW near-term connection pipeline. Earlier Dealroom coverage tracked the compute launch and Fuse's move into generation and supply. This post is the clearest public statement yet of why gas generation sits next to that compute stack: firm power that AI loads can soak up around the clock.
Hiring. Chang is recruiting for Fuse's projects team ("the build-everything team"), looking for electrical and mechanical engineers who want to build without politics and red tape.
Sources: Alan Chang on X; Fuse Compute.
Photo: Alan Chang, co-founder and CEO of Fuse Energy (London Tech Week).