Analysis
Etched closed a $300 million Series C at a $10.3 billion valuation, led by Sequoia Capital with Andreessen Horowitz, SK Hynix, Jane Street, and Diffusion Capital joining existing backers. The round roughly doubles the company's valuation in just seven months, after Etched raised $500 million at a $5 billion valuation in December 2025. Etched says it has now booked more than $1 billion in orders for Sohu, its inference-only chip, and has already manufactured its first units.
Founded in 2022 by Harvard dropouts Gavin Uberti, Chris Zhu, and Robert Wachen, Etched bet against the industry consensus that general-purpose GPUs would keep winning. Instead of building flexible chips like Nvidia's H100 or B200, Etched hard-wired its silicon around the transformer architecture that underpins ChatGPT, Claude, and Gemini, trading flexibility for a claimed multiple-fold efficiency gain on inference workloads specifically.
“Founded in 2022 by Harvard dropouts Gavin Uberti, Chris Zhu, and Robert Wachen, Etched bet against the industry consensus that general-purpose GPUs would keep winning.”
The competitive landscape has gotten crowded fast: Groq, Cerebras, SambaNova, and Positron are all chasing pieces of the inference market as compute spending shifts from training frontier models to serving them cheaply at scale. Nvidia still commands the overwhelming majority of AI accelerator revenue, but critics argue the inference era rewards specialization over generality, exactly Etched's pitch to investors.
A $10.3 billion valuation for a Series C is unusual even in this market, but it lands in the same week as several other AI infrastructure megadeals, including Together AI's $800 million raise earlier this month and SambaNova's $1 billion round. What to watch: whether Etched can convert its $1 billion order book into shipped, revenue-generating hardware at the volumes its valuation implies, and whether Nvidia's own inference-optimized roadmap narrows the efficiency gap Etched is selling.
