Etched
AI Chip Startup
About Etched
Etched is an AI semiconductor company building purpose-built hardware for large-scale AI inference. Founded in 2022 by Harvard dropouts Gavin Uberti, Chris Zhu, and Robert Wachen, the company develops custom AI chips and rack-scale inference systems designed to accelerate modern foundation models.
Rather than relying on general-purpose GPUs, Etched’s architecture is optimized specifically for inference workloads, enabling significantly higher throughput and improved energy efficiency for serving large language models in production.
Etched’s flagship platform combines custom silicon with proprietary networking, memory architecture, and compiler software to maximize inference performance at the data center level. After emerging from stealth in 2026, the company announced working production silicon, over $1 billion in signed customer contracts, and the testing of its first full-scale systems by customers.
Founders
Funding History
Frequently Asked Questions
What does Etched build?
Etched develops specialized AI inference hardware, including custom chips, rack-scale systems and software optimized for running transformer and mixture-of-experts models with high throughput and energy efficiency.
When was Etched founded?
Etched was founded in 2022 by Gavin Uberti, Chris Zhu and Robert Wachen.
How much funding has Etched raised?
Etched has publicly disclosed total funding of approximately $925.46 million across Seed, Grant, Series A, Series B and Series C financing rounds.
What is Etched's latest valuation?
Etched's latest publicly announced valuation is $10.3 billion following its $300 million Series C funding round announced on 23 July 2026.
Who are Etched's major investors?
Investors include Sequoia Capital, Andreessen Horowitz, SK hynix, Jane Street, Peter Thiel, Primary Venture Partners, Positive Sum, Ribbit Capital, Stripes, Radical Ventures, Argo and Diffusion Capital Partners, among others.
Where is Etched headquartered?
Etched is headquartered in San Jose, California, United States.
