Etched, the AI chip startup founded by ex-Cerebras engineers, has doubled its valuation to $21 billion in roughly one month. Jane Street, the quantitative trading firm, deployed Etched's first shipped AI cluster system and subsequently led a new funding round based on the results.
The rapid valuation jump reflects investor confidence in Etched's custom silicon approach to AI inference. The startup builds processors optimized for specific AI workloads rather than general-purpose GPUs. Jane Street's deployment and decision to lead the follow-on round signals that Etched's technology delivers measurable performance advantages in production environments.
Etched competes directly against Nvidia and startups like Cerebras and Groq in the race to own AI infrastructure. While Nvidia dominates with its existing market position and CUDA ecosystem, companies like Etched argue that custom silicon can deliver better price-to-performance ratios for specific tasks. Jane Street's endorsement matters because the firm runs computationally intensive operations and has resources to evaluate multiple solutions. Their choice to lead funding after testing the hardware carries credibility beyond typical investor interest.
The valuation jump in a month is extreme, even for AI infrastructure. It suggests either a massive round size or significant upward revision from prior funding. Etched likely closed its Series B or C at a substantially higher valuation than previous rounds, driven by hard shipping evidence and a marquee customer signing up to use the technology at scale.
For Etched's founders, the timing proves crucial. The AI infrastructure market remains young. Early evidence of product-market fit, customer traction, and superior performance can unlock venture capital in compressed timelines. Jane Street's participation as both customer and lead investor removes execution risk from Etched's perspective and validates the core thesis that custom silicon beats general-purpose alternatives for certain workloads.
The bigger question for Et
