Etched AI Chip Startup Secures Over $1 B in Inference Chip Orders

Etched, a four‑year‑old U.S. artificial‑intelligence chip startup, announced that it has received more than $1 billion in orders for hardware designed specifically for AI inference – the stage where trained models generate outputs. The company says its "frontier inference clusters" combine custom silicon, networking and software, and are aimed at reducing the cost and energy use of running large language models compared with Nvidia’s general‑purpose GPUs.

Etched disclosed a valuation of $5 billion and is preparing a new financing round that could lift its valuation to around $20 billion, with a $10 billion‑valued round led by Sequoia Capital already under way. Its first chip is being fabricated by Taiwan Semiconductor Manufacturing Company (TSMC). The firm employs roughly 400 engineers recruited from Nvidia, Google’s TPU team, Broadcom, SK Hynix and TSMC, and hopes the incoming orders will divert a portion of the market currently dominated by Nvidia.

The startup’s co‑founder and CEO Gavin Uberti told TechCrunch that specialized inference chips are "inevitable" as the economics of deploying frontier AI models evolve.