Skip to content
NewsFunding & M&A

Etched raises $700 million at $21 billion valuation as Jane Street puts first inference rack into production

· by Pondero Newsdesk

The short version

Etched announced $700 million in new funding at a $21 billion valuation on August 18, with Jane Street as lead investor and first paying customer after deploying the company's inference chips in its own data center.

Etched raises $700 million at $21 billion valuation as Jane Street puts first inference rack into production

Jane Street is running Etched hardware in production. That fact, disclosed alongside the $700 million raise announced August 18, shifts the question around this inference chip startup from "will it ever ship" to "who else can get access."

What happened

Per Etched's official press release via GlobeNewswire, the company raised $700 million in a round led by Jane Street, with participation from Kleiner Perkins, Sequoia, Andreessen Horowitz, Tiger Global, Bain Capital Ventures, Neo, Primary, Stripes, Positive Sum, and Blackstone. The round values Etched at $21 billion.

The valuation doubled in under 30 days. Etched closed its Series C at $10.3 billion on July 23, less than four weeks before this announcement. Total capital raised across all rounds now stands at $1.9 billion, per the same release.

Alongside the funding, Etched confirmed it shipped its first commercial rack to Jane Street last month. The firm is running the hardware in its own data center. Jane Street said in a statement included in the press release: "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."

How the hardware works

Etched builds purpose-built inference chips rather than general-purpose GPU-style hardware. Where a GPU handles many compute workload types through programmable shader cores, Etched's chips implement inference operations as fixed silicon circuits optimized for a single task: running AI models as fast as possible at the lowest possible power draw. The company calls this architectural approach transformer-specific inference, and the SOHU chip is the first production expression of it.

Two core technologies define the performance architecture. Low Voltage Inference delivers higher compute density within the same power envelope as existing hardware. Cluster Scale Memory pools memory across an entire cluster rather than limiting it to a single chip's buffer. The press release states the clusters are already running "massive MoE models and non-transformer designs." Etched also disclosed more than $1 billion in signed customer contracts from "public and private frontier AI companies and clouds."

Why it matters

Etched achieved first-pass silicon success in under three years from seed funding. The company emerged from stealth in June 2026 with more than 400 employees and a working chip, closed its Series C weeks later at what the press release called the highest valuation ever for a Sequoia-led round, and now doubled again with a live customer deployment to show for it. That pace is unusual for a hardware company taking on GPU incumbents.

Jane Street's involvement is not a passive portfolio bet. The firm led this round and simultaneously deployed the hardware into production. For enterprise buyers evaluating purpose-built inference alternatives to GPU clouds, a named production deployment at one of the world's most rigorous quantitative trading firms carries weight that benchmark marketing alone does not.

At $21 billion, Etched is large enough that its trajectory shapes the inference market regardless of whether individual developers can access its chips today. Nvidia has not made a public throughput-per-watt response to Etched's claims. Whether that silence holds as more customer deployments become public will define the competitive framing through the rest of 2026.

What to watch next

Three questions will determine Etched's next chapter. Whether the company announces a cloud or hyperscaler partnership that opens its chips to developers outside its direct contract book. Whether additional named customers appear beyond Jane Street. And whether Nvidia or another incumbent responds publicly to the inference throughput and power efficiency claims.

Sources