AI inference hardware startup Etched has more than doubled its valuation to approximately $21 billion in roughly one month, closing a fresh $700 million funding round led by quantitative trading giant Jane Street, according to reports from The Wall Street Journal and BitcoinWorld on August 18, 2026.
The new valuation follows a rapid climb for the company: it was valued at $5 billion in December 2025 and $10.3 billion after a Sequoia-led $300 million Series C round in late July 2026. The latest financing also included participation from Andreessen Horowitz, SK Hynix, Diffusion Capital and others.
Jane Street has moved beyond investing to become Etched's first customer. The Wall Street firm deployed a server rack filled with Etched's AI processors optimized for inference computing. In a blog post, Jane Street said, "We tested the chip and are pleased with the early results. Etched's unique approach to inference delivers the precision we will need to support our most demanding workloads."
Etched develops full rack systems focused on the inference portion of AI compute, which occurs after users submit prompts. Its architecture splits work into a low-voltage prefill phase and a decode phase using cluster-scale memory, a design the company says allows chips to run cooler, share memory across accelerators, and deliver higher speeds at lower costs. Etched says its systems are already inside DeepSeek, Qwen, Mamba and Llama models.
The funding underscores the broader rush for alternatives to Nvidia's dominance in AI silicon. Bloomberg Intelligence has projected the inference market could reach $1.3 trillion, doubling the AI training market by 2032, while Structure Research expects inference to account for 80% of AI compute load by 2030. Etched reported more than $1 billion in booked orders, employs over 400 people and operates data center capacity in San Jose and Milpitas after achieving first-pass silicon success on TSMC's N4P process.
