A startup founded by two Harvard dropouts just crossed the $5 billion valuation mark, and it did so by making a very specific bet: that the future of AI hardware isn’t general-purpose GPUs, but purpose-built chips that do one thing extremely well.
Etched, based in San Jose, has booked over $1B in forward contracts for inference systems powered by its custom chip called Sohu. The company raised roughly $500M in a funding round led by Stripes in January 2026, bringing its total capital raised to approximately $800M.
What Sohu actually does
Instead of building a chip that can handle everything from gaming to scientific simulation to AI training, Etched built Sohu as a custom ASIC, an application-specific integrated circuit, designed to do exactly one thing: run transformer model inference as fast as physically possible.
The performance claims are aggressive. Etched says a single 8-chip Sohu server can process around 500,000 tokens per second running Meta’s Llama 70B model. The company claims that setup outperforms 160 Nvidia H100 GPUs while using less power and taking up less physical space.
