Jane Street's Live Etched Cluster Sends Inference ASIC Valuations Into Orbit
Etched hits $21B in one month after Jane Street deploys its first frontier inference cluster, signaling institutional conviction in ASIC inference hardware.
2. Jane Street's Live Etched Cluster Sends Inference ASIC Valuations Into Orbit
Etched announced on August 18, 2026, that it raised $700 million at a $21 billion valuation, led by Jane Street. The round follows a $300 million Series C at a $10.3 billion valuation in July and a $5 billion valuation in December 2025. The step-up: nearly $11 billion in roughly one month. Jane Street did not just write a check. The quant fund installed an Etched frontier inference cluster in its own datacenter, tested it, and then led the round. The hardware centers on two purpose-built components: a low-voltage prefill chip that packs more transistors without heat penalties, and a cluster-scale memory interconnect that lets many chips share a unified memory pool at low latency for the decode phase.
The strategic read here is about who now holds pricing power in inference silicon. Nvidia sells what it calls AI factories. Etched sells frontier inference clusters and is attacking the same category from a different architecture. Where Nvidia's GPUs are general-purpose, Etched's chips are optimized specifically for the two-stage inference pipeline: prefill and decode. Jane Street's decision to deploy before leading the round is the tell. Quant firms do not install unproven hardware for optics. That one real-world deployment converted a speculative hardware bet into a production endorsement, and the market repriced Etched accordingly. Kleiner Perkins, Sequoia, Andreessen Horowitz, Blackstone, and Peter Thiel are all already in.
The broader pattern is institutional capital moving from model bets toward infrastructure bets. Etched's valuation arc, $5 billion to $21 billion in eight months, tracks the moment when inference cost became the dominant AI operating concern. Watch whether other quant funds or hyperscalers follow Jane Street's move and begin deploying ASIC inference clusters as a cost-reduction play against GPU-based inference at scale.