Low-power AI inference chips
axelera.ai (opens in a new tab)From its headquarters on Eindhoven's High Tech Campus, Axelera AI designs silicon for AI inference. One low-power architecture covers everything from embedded devices at the edge to large inference infrastructure, with performance per watt as the aim. Edge inference platforms are shipping, and a next-generation architecture is built specifically for the economics of inference at scale. In the US, where the company is expanding its commercial team, the target buyers are those most squeezed by power limits and total cost of ownership: frontier model developers, neoclouds and hyperscalers. It employs more than 250 people, over 60 of them PhDs whose work has drawn over 40,000 citations.