AI chip startup Etched has quietly closed one of the largest private raises in the AI hardware race. The company has secured nearly $500 million in fresh funding as it sharpens its challenge to Nvidia. The round values the company at nearly $5 billion and pushes its total funding close to the $1 billion mark. This is a notable development for a company, which is preparing it’s first major chip.
The latest funding round was led by Stripes, with participation from Peter Thiel, Positive Sum, and Ribbit Capital. Earlier supporters include Primary Venture Partners, along with notable angels such as GitHub CEO Thomas Dohmke and former Coinbase executive Balaji Srinivasan.
A bold bet on purpose-built silicon
Etched was founded in 2022 by Chris Zhu, Gavin Uberti, and Robert Wachen, who dropped out of Harvard with a focused goal. They wanted to build chips designed specifically for transformer-based AI models. They were later joined by CTO Mark Ross, who rethinks AI hardware from the silicon up.
At the centre of that effort is Sohu, Etched’s custom AI accelerator. Instead of chasing broad workloads, the chip is engineered for a narrow but exploding category of AI computation.
To bring it to life, Etched partnered with Taiwan Semiconductor Manufacturing Co.’s Emerging Businesses Group, a move that signals both technical ambition and manufacturing credibility. The team has also drawn talent from established chipmakers such as Cypress Semiconductor and Broadcom, blending startup speed with industry experience.
Competing against Nvidia
The challenge Etched faces is formidable. Nvidia’s dominance in AI accelerators remains unmatched, and the company recently projected more than $500 billion in cumulative data-centre sales by the end of 2026.
Yet Etched is not trying to replace Nvidia everywhere. Its strategy hinges on doing one thing exceptionally well, which is running transformer models more efficiently than general-purpose GPUs. If Sohu delivers on that promise, Etched could carve out a valuable niche in an overcrowded but rapidly expanding market, proving that focus, not scale alone, can still move the needle in AI hardware.
