Summary: Etched said on June 30, 2026 that it has $1 billion in contracted orders, produced its first chip at TSMC earlier in the year, and sent Frontier inference clusters to customers for testing. The complete systems combine Etched chips, custom racks, and software. The company has not named those customers or published exact chip specifications.
On June 30, 2026, AI chip startup Etched put out a progress report under the banner “Out of stealth,” and the contents are anything but vague: it holds $1 billion in orders under contract, its first chip came off TSMC’s lines earlier this year, and the Frontier inference cluster — chips plus custom racks and software as one complete setup — is now with customers for testing.
Started in 2022 by two founders who left Harvard — CEO Gavin Uberti and President Robert Wachen, both Thiel Fellows — the company has talked to reporters since 2024; until now it had never shown its operating figures and product status all at once.
Surviving the Early Years, Then Landing $1 Billion in Orders
Etched nearly never made it out of the gate. In 2023 the money in the bank was almost gone. The founders had written a 30-page memo arguing that AI would ultimately need specialized chips instead of general-purpose GPUs, yet not one major investor went for it — in a market still wedded to Nvidia’s general-purpose accelerators as the answer to every workload, the idea read as simply too contrarian. By 2024, cumulative funding stood at just over $125 million, a slim war chest in a field where a single advanced-node tape-out can burn through hundreds of millions of dollars.
The break came in December 2025: Stripes led a $500 million raise completed without any public word, valuing the company at $5 billion post-money and taking total funding to around $800 million. That valuation shows how quickly sentiment turned — the thesis investors passed on in 2023 had become one of semiconductor investing’s most contested bets. The $1 billion now booked as contracts takes the story further: buyers have moved past the pitch and are placing real orders.
Frontier Inference Clusters: Selling the Complete System, Not a Chip
Loose chips are not the offering; the core product is the Frontier inference cluster: the company’s own chips plus custom racks and software, packaged into whole systems that, the company claims, beat rivals on speed, cost, and power efficiency at inference. The product is still in customer testing, and the update stays quiet on two things: who those customers are, and what the chip’s exact specifications are. What the record confirms is that the chip was successfully manufactured at TSMC earlier this year, so the obstacle that kills many a fabless startup — getting silicon made at all — already sits behind Etched.
The positioning follows from how the market actually buys: purchasers want inference capacity ready to drop straight into production, and a lone chip does not deliver that. Nearly every startup in Etched’s cohort has adopted that same “system plus service” formula, turning hardware-level efficiency into compute sold by the unit. Buyers, in turn, will judge on delivered cost per token instead of spec sheets.
A Cap Table Loaded With Big Names
Stripes leads the round; alongside it sit VentureTech Alliance, quant-trading heavyweights Jane Street, Hudson River Trading, and Two Sigma, together with Ribbit Capital. The individual investors are the rare part: Andrej Karpathy, Geoffrey Hinton, Fei-Fei Li, Arthur Mensch, and Scott Wu are each named, with billionaires Stanley Druckenmiller and Peter Thiel also on the list.
One part of that list deserves a second look: the quant funds. Heavy buyers of inference compute themselves, the funds gain insurance for their own supply chain by holding a stake in Etched — and the leading AI researchers alongside them amount to an endorsement of the specialized-inference-hardware thesis.
The Field It Is Walking Into Is Already Full
Etched surfaces at the peak of the inference-chip capital cycle. Cerebras wrapped up the year’s standout IPO in May, pulling in $5.5 billion; Groq confirmed on June 22 that it had secured $650 million. The cloud giants are deep into their own chip efforts — Amazon, Google, and Microsoft all run AI chip programs — and OpenAI showed its first custom chip, developed with Broadcom, on June 24.
In short, “beyond Nvidia” has gone from slogan to real racetrack — and on it, few independent players can show both a $1 billion order book and successfully manufactured silicon, Etched among them.
Why Now Is Worth Watching
For compute buyers, each Etched advance further diversifies inference supply; once specialized systems’ efficiency claims hold up one by one, inference unit costs gain room to fall. The truly demanding test is only starting — between customer testing and deployment at scale stand yields, supply chains, and the day-to-day of running live data centers. The $1 billion order book marks a milestone, not the destination; the signal to watch next is which customers shift production inference workloads over to Frontier clusters.