Etched raises $700M at $21B, doubling its valuation in a month

Etched said on Tuesday that it raised another $700 million at a $21 billion valuation, in a round led by Jane Street. The announcement came weeks after the inference chip startup closed a $300 million Series C at $10.3 billion, backed by Nvidia. The Next Web reports the new raise as a Series D.

Even by AI standards, that pace is unusual. TechCrunch notes the company was valued at $5 billion in December, which means investors have roughly quadrupled its price in eight months, by our count. Kleiner Perkins, Sequoia, Andreessen Horowitz, Tiger Global and Bain Capital Ventures joined the round, alongside several others.

Date	Valuation	What happened
December 2025	$5 billion	Prior mark, per TechCrunch
July 2026	$10.3 billion	$300 million Series C
August 2026	$21 billion	$700 million round led by Jane Street

Why a quant fund led the round

Jane Street isn’t just the lead investor. It’s also Etched’s first customer, and it installed an Etched rack in its own data center last month after testing the chip. That sequence, try the hardware and then price the company, is what the firm itself described.

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. We’re excited to now have our own rack running in our datacenter.

Jane Street, via Etched’s funding announcement

Etched says it has signed more than $1 billion in customer contracts. It also built a 2-megawatt AI cluster at its San Jose headquarters, which prospective customers use to test its chips, per SiliconANGLE. The market hasn’t treated every inference chip maker this way, though. We covered how Groq raised $350 million at $3.5 billion, down from $6.9 billion.

What the chips actually do

Inference is the computing that happens after you submit a prompt, and it splits into two stages. Co-founder and COO Robert Wachen told TechCrunch that the compute-heavy prefill stage digests the prompt, while the memory-hungry decode stage generates the answer. Etched designed a new component for each.

The prefill chip runs at low voltage, which lets it pack in more transistors without the heat problems typical of high-end AI chips. Because of that, TechCrunch reports, it can process more tokens faster. Etched says the chip’s math blocks run at under half the voltage of most AI accelerators, and claims “multiple times the FLOPs density” of rivals, per SiliconANGLE. Power efficiency carries extra weight now that the AI compute shortage has moved from chips to electricity.

For decode, the company built new memory and an interconnect it calls cluster-scale memory. “It allows many chips to connect together and use a shared memory pool at a very, very fast, low latency,” Wachen said. SiliconANGLE, citing the Wall Street Journal, reports the in-house interconnect finishing some communication tasks in 700 milliseconds that take rival chips 4,000 milliseconds. The result, the company promises, is higher speeds and lower costs.

Etched launched in 2021 planning chips tied to specific AI models, then pivoted to accelerators that run a variety of architectures. TechCrunch notes it’s still fighting the old perception that each chip is etched for one model, and says its systems can run any frontier model. Its first prototype came off a TSMC production line earlier this year, and Jane Street signed on shortly after.

“It took us three years to deliver our first rack from scratch,” CEO Gavin Uberti said, per SiliconANGLE. “Our next one will be much faster.” The valuation is the headline number, but delivery pace is the real test. Etched sells full systems it calls frontier inference clusters, and it’s up against Nvidia, which brands its own as AI factories.