Inference Chip Startup Etched Raises $700M at $21B Valuation

The new funding more than doubles the company’s valuation. Etched was founded in 2021 with the goal of designing chips optimized for specific artificial intelligence models, and later pivoted to making inference accelerators that support a variety of model architectures. The company said its first prototype chip rolled off a Taiwan Semiconductor Manufacturing Co. production line earlier this year. Jane Street, the lead investor in the current round, became Etched’s first customer shortly after.

Jane Street installed an Etched-powered rack in its data center last month. Etched also built a two-megawatt AI cluster at its San Jose, California headquarters, which prospective customers use to test the company’s chips.

Etched says its chip’s math blocks, or math-optimized circuits, run at under half the voltage of most AI accelerators, allowing the processor to provide “multiple times the FLOPs density” offered by rivals. FLOPs density is a measure of a processor’s performance. The company did not disclose how the voltage-optimized math blocks work. Large language models perform two main types of mathematical operations during inference — dot product calculations and matrix multiplications — and the blocks are presumably optimized for those tasks.

Etched ships its silicon as part of custom-designed racks. One of the design’s main selling points is an internally developed interconnect that links together the chips in each appliance. According to the Wall Street Journal, the interconnect can complete some communications tasks that take rival chips 4,000 milliseconds in 700 milliseconds. The appliances also contain custom cold plates, which conduct heat from the chips to coolant, and a proprietary voltage regulator module that optimizes the flow of electricity.

“It took us three years to deliver our first rack from scratch. Our next one will be much faster,” said Etched co-founder and Chief Executive Officer Gavin Uberti.