Etched, the AI inference chip startup founded by three Harvard dropouts, has raised $800 million led by trading firm Jane Street and TSMC-linked VC firms, with Jane Street alone having invested over $100 million total. The company is simultaneously raising capital at two valuations: $10.3 billion in a Series C led by Sequoia Capital, and at a $20 billion target valuation in Jane Street's round, which would quadruple its prior $5 billion valuation.
Etched has $1 billion in pre-commercial customer demand locked in despite having not yet commercially shipped its first product, signaling exceptional procurement confidence from enterprises planning AI inference capacity. The company designed two novel components for AI inference: a prefill chip that runs at dramatically lower voltage to reduce heat and pack more transistors, and cluster-scale memory interconnect technology for high-speed, low-latency multi-chip communication. These optimizations target the economics of inference workloads where training GPUs are mismatched to deployed model requirements.
For data center architects, Etched's $1 billion pre-release demand backlog is the clearest signal that procurement teams are reserving capacity in specialized inference silicon now, months before product availability, rather than waiting for proven supply. The back-to-back valuation rounds signal investor conviction that the inference-silicon market is durable and competitive enough to support new chipmakers beyond NVIDIA, Cerebras, and Groq.