Etched has silicon, a customer rack running, and $1bn raised in 26 days
The first customer is also the lead investor, which blunts the validation. What is left is a non-Nvidia inference system buyers can benchmark instead of read about.
The Product Desk
What happened
- Etched announced on Tuesday that it raised another $700 million at a $21 billion valuation, led by Jane Street after the quant fund tested and bought the startup's AI hardware.
- The raise more than doubles Etched's valuation to $21 billion; Jane Street was joined by Kleiner Perkins, Sequoia, Andreessen Horowitz, Tiger Global, Bain Capital Ventures and several others.
- Etched was valued at $5 billion in December and raised a $300 million Series C at a $10.3 billion valuation in July.
- The $700 million round came a few weeks after Etched closed a $300 million round backed by Nvidia Corp.
- The valuation doubled in 26 days: Etched was worth $10.3 billion on 23 July after a $300 million Series C led by Sequoia, and $5 billion in December, roughly four times in eight months, at a company that has just made its first delivery to a paying customer.
Why it matters
Etched announced on Tuesday that it raised $700 million at a $21 billion valuation in a round led by Jane Street, the quant fund that tested and bought its hardware [1]. That matters less for the number than for what sits behind it: a chip out of TSMC, a rack running in a paying customer's data center, and a second option for anyone whose inference budget is currently a single-vendor line item [8][6].
The ladder is steep. Etched was worth $5 billion in December and $10.3 billion on 23 July after a $300 million Series C led by Sequoia, a round Nvidia also backed [3][5][4]. The new raise doubled that in 26 days, roughly four times in eight months [5][2]. Combined, the two rounds put about $1 billion into the company inside that window [17].
What is physically true is narrower than the valuation implies, and more interesting. Etched says its first prototype came off a TSMC line earlier this year [8], working on the N4P process at the first attempt, which TNW notes is uncommon [9]. Jane Street installed an Etched-powered rack in its data center last month and said in Jane Street's own blog post that it tested the chip and is "pleased with the early results" [6][7]. Etched has also built a two-megawatt cluster at its San Jose headquarters for prospective customers to run their own tests [11]. That is the part a buyer can act on: a benchmarking target that exists.
The design splits inference. COO Robert Wachen told TechCrunch that a low-voltage prefill chip packs in more transistors without the usual heat ceiling, while decode runs on an interconnect and memory scheme the company calls cluster-scale memory, letting many chips share one fast pool [10]. Etched says its math blocks run at under half the voltage of most AI accelerators, yielding "multiple times the FLOPs density" of rivals, without disclosing how [23]. SiliconANGLE, citing the Wall Street Journal, reports the interconnect completes some communications tasks in 700 milliseconds that take rival chips 4,000 [24], about a 5.7x improvement on a single undefined task [25]. The racks also carry in-house cold plates and a proprietary voltage regulator module [12]. Crucially for procurement, the one-model-per-chip premise is gone: Etched says its systems run Mixture of Experts models including DeepSeek and Qwen, plus non-transformer designs such as Mamba [13][22].
The discount to apply is the customer list. TNW reported in June that Jane Street was already a backer, alongside Hudson River Trading, Jump Trading and Two Sigma, so this is an existing investor increasing its position and taking delivery rather than an arm's-length buyer converting [14]. Contracts stood at more than $1 billion as of June [15], which puts the valuation at roughly 21 times the signed book [21].
Watch three things. Whether a customer with no equity stake takes delivery and says something specific about tokens per dollar. Whether the build cadence changes: CEO Gavin Uberti says the first rack took three years and the next will be "much faster" [20]. And where the money goes, since Etched says the round funds factories, supply chains and fleet software toward gigawatt scale [18], the exact stage where accelerator startups usually stall. London's Olix recently tripled to $3.3 billion chasing the same market [19].
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence55
- Adoption20
- Hype gap+40
- Incentives70
- Confidence60
Perspective Coverage
4 publishers- Builder
- Builder 34%
- Operator
- Operator 24%
- Investor
- Investor 42%
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Etched announced on Tuesday that it raised another $700 million at a $21 billion valuation, led by Jane Street after the quant fund tested and bought the startup's AI hardware.
- [2]
The raise more than doubles Etched's valuation to $21 billion; Jane Street was joined by Kleiner Perkins, Sequoia, Andreessen Horowitz, Tiger Global, Bain Capital Ventures and several others.
- [3]
Etched was valued at $5 billion in December and raised a $300 million Series C at a $10.3 billion valuation in July.
- [4]
The $700 million round came a few weeks after Etched closed a $300 million round backed by Nvidia Corp.
- [5]
The valuation doubled in 26 days: Etched was worth $10.3 billion on 23 July after a $300 million Series C led by Sequoia, and $5 billion in December, roughly four times in eight months, at a company that has just made its first delivery to a paying customer.
- [6]
Etched signed Jane Street as its first customer shortly after the prototype came off the TSMC line, and the investment firm installed an Etched-powered rack in its data center last month.
- [7]
In the blog post announcing the round, 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. We're excited to now have our own rack running in our datacenter."
- [8]
According to Etched, its first prototype chip rolled off a Taiwan Semiconductor Manufacturing Co. production line earlier this year.
- [9]
Etched's first chip worked on TSMC's N4P process at the first attempt, which is not common.
- [10]
Co-founder and COO Robert Wachen told TechCrunch that inference is built in two stages, prefill and decode; Etched created a prefill chip that operates at low voltage, allowing it to pack in more transistors without the typical heat problems of other high-end AI chips, and a new memory and interconnect for decode that it calls cluster-scale memory, allowing many chips to share a fast, low-latency memory pool.
- [11]
Etched built a two-megawatt AI cluster at its San Jose, California headquarters, which prospective customers use to test its chips.
- [12]
Etched's appliances contain internally developed custom cold plates that conduct chip heat to the circulated coolant, and a proprietary voltage regulator module that optimizes electricity flow.
- [13]
Etched was founded on etching a single model into silicon and now says its systems run Mixture of Experts designs including DeepSeek and Qwen, and non-transformer architectures such as Mamba.
- [14]
TNW reported in June that Jane Street was already among Etched's backers, alongside Hudson River Trading, Jump Trading and Two Sigma, so the round is an existing investor increasing its position and taking delivery rather than a customer converting into a lead.
- [15]
More than $1 billion in contracts had been signed as of June, a commercial base TNW describes as real but thin.
- [16]
Etched delivers its technology as full systems it calls "frontier inference clusters"; Nvidia calls its full systems AI factories.
- [17]
The $300 million Series C and the $700 million round together total about $1 billion raised within the 26-day window between the two valuations.
- [18]
Etched says the round funds factories, supply chains and fleet software on the way to gigawatt scale.
- [19]
European challengers are chasing the same market, with London's Olix recently tripling to a $3.3 billion valuation.
- [20]
"It took us three years to deliver our first rack from scratch," said Etched co-founder and CEO Gavin Uberti. "Our next one will be much faster."
- [21]
The $21 billion valuation is roughly 21 times the reported contract book of more than $1 billion.
- [22]
Etched launched in 2021 aiming to design chips optimized for specific AI models, then pivoted to inference accelerators that support a variety of model architectures.
- [23]
Etched says its chip's math blocks run at under half the voltage of most AI accelerators, so the processor can provide "multiple times the FLOPs density" of rivals; Etched did not disclose how the voltage-optimized math blocks work.
- [24]
According to the Wall Street Journal, as reported by SiliconANGLE, Etched's internally developed interconnect can complete some communications tasks that take rival chips 4,000 milliseconds in 700 milliseconds.
- [25]
The cited interconnect figures represent roughly a 5.7-fold reduction in time on the unspecified task.
Sources
4 independent publishers whose own reporting we read for this story.
- datacenterdynamics.comInference chip startup Etched raises $700m, doubles valuation to $21bn
1 article · August 19, 2026
- siliconangle.comInference chip startup Etched raises another $700M at $21B valuation
1 article · August 18, 2026
- techcrunch.comEtched’s valuation doubles to $21B in a month
1 article · August 18, 2026
- thenextweb.comEtched raises $700M at a $21B valuation led by Jane Street
1 article · August 18, 2026