Etched, an AI inference chip startup, raised $300 million in Series C funding at a $10.3 billion valuation, led by Sequoia Capital with participation from Andreessen Horowitz, Jane Street, SK Hynix, and Diffusion Capital. The round represents the highest valuation ever for a Sequoia-led Series C. Etched said it has booked over $1 billion in customer pre-orders for its rack-scale inference systems.
The company, which emerged from stealth less than a month ago, is using the funding to scale production and customer deployments. It opened a new 80,000-square-foot facility near San Jose with a 10 MW data center plus an NPI lab and in-house SMT line. Etched has also established a Taiwan factory and is ramping production to fulfill demand as customers transition from evaluation to deployment phases.
Etched's architecture targets inference bottlenecks through two novel computing technologies: low-voltage inference (LVI) to reduce thermal throttling and maximize utilization, and system-level optimization from chip design through cooling and mechanical engineering. The company runs its inference math engines at under half the voltage of competing AI chips, enabling 80%+ utilization for trillion-parameter models without thermal throttling.
For infrastructure buyers, Etched's $1B backlog signals real procurement momentum at hyperscale providers despite the chip being unvalidated independently. The company's doubled valuation in seven months (from $5B in December) reflects investor conviction that purpose-built inference compute, not general-purpose GPUs, will power the majority of the world's inference workloads.