What is Etched?

Etched went from receiving silicon to running inference in a rack in 40 days, versus 10 months for a well-known AI chip competitor.

“There was another very famous AI chip company that took 10 months to go from getting their silicon back to having them running inference in Iraq. And this was like publicly announced to their investors and was it was it was a really big deal. we were able to do it in 40 days.”

Etched. A startup building transformer-specific ASICs targeting inference. Claims 10x concurrency vs. NVIDIA, sub-40-day silicon bring-up, and a novel low-voltage process. Raised $800M and secured early TSMC cooperation before its Series A.

“Our hardware is going to generally be able to get you an order of magnitude more concurrency at a given level of interactivity.”

Jake Paul's growth fund holds positions in SpaceX, OpenAI, Anthropic, Andal, Etched, and Cognition, spanning frontier AI and compute.

“A lot of the tier one names, maybe all the tier one names. but andal, etched, cognition, spacex, open AI, anthropic.”

Gavin Uberti reframing the Etched vs. NVIDIA compute question as additive, not zero-sum.

“It's not a decision between a gigawatt of a GPU and a gigawatt of us. It's 2 gawatt.”

Every major Silicon Valley investor passed immediately on Etched's Series A, despite the company later raising $800M.

“Every major investor in the valley passed immediately.”

Etched claims its hardware delivers an order of magnitude more concurrency than NVIDIA at equivalent interactivity levels.

“Our hardware is going to generally be able to get you an order of magnitude more concurrency at a given level of interactivity.”

Etched's low-voltage inference technology runs at under half the voltage of any other AI chip on the market.

“Running at under half the voltage of any other.”