On June 30, 2026, AI chip startup Etched published a progress report and declared itself “out of stealth”: the company holds $1 billion in booked orders under contract, its first chip was successfully manufactured by TSMC earlier this year, and its Frontier inference clusters — full systems bundling the chips with custom racks and software — are now in customer testing.
Founded in 2022 by two Harvard dropouts, CEO Gavin Uberti and President Robert Wachen, both Thiel Fellows, the company had spoken with reporters since 2024. But this is the first time it has laid out its operating numbers and product status in full.
From Near-Death to $1 Billion in Orders
Etched’s opening chapter almost ended in failure. In 2023 the company nearly ran out of cash. Its founders had written a 30-page memo arguing that AI would eventually need specialized chips rather than general-purpose GPUs — and every major investor passed. The thesis was too contrarian for a market still treating Nvidia’s general-purpose accelerators as the default answer to every workload. By 2024, the company had raised just over $125 million, a modest war chest by AI silicon standards, where a single advanced-node tape-out can consume hundreds of millions of dollars.
The turn came in December 2025, when an unannounced $500 million round closed at a $5 billion post-money valuation, led by Stripes. Counting that round, Etched has raised roughly $800 million in total. The valuation alone tells the story of how quickly sentiment shifted: investors who declined in 2023 watched the same thesis become one of the most contested bets in semiconductors. And the $1 billion in booked orders signals that the market has moved on from listening to the story — customers are now placing real contracts against it.
Frontier Inference Clusters: Selling Systems, Not Just Chips
The product shape deserves attention. Etched does not sell bare silicon; it sells Frontier inference clusters, packaging its chips with custom racks and software into complete systems that it claims run inference faster, cheaper, and more power-efficiently than rivals. The product is still in customer testing, and the company has not named its customers or disclosed exact chip specifications in this update — what it has confirmed is that TSMC successfully manufactured the chip earlier this year, clearing the manufacturing hurdle that sinks many fabless startups.
The positioning reflects a market reality: buyers want production-ready inference capacity, not a chip. Most of Etched’s peer startups have converged on the same “system plus service” model, translating hardware efficiency into purchasable units of compute. For customers, that shifts the comparison from spec sheets to delivered cost per token.
A Striking Investor Roster
Beyond lead investor Stripes, the cap table includes VentureTech Alliance and quantitative trading giants Jane Street, Hudson River Trading, and Two Sigma, plus Ribbit Capital. The angel list is unusually strong: Andrej Karpathy, Geoffrey Hinton, Fei-Fei Li, Arthur Mensch, and Scott Wu are all named investors, and billionaires Stanley Druckenmiller and Peter Thiel hold stakes as well.
The quant funds are a telling detail. They are themselves heavy buyers of inference compute, so investing doubles as insurance on their own supply chain — while the presence of top AI researchers reads as an endorsement of the specialized-inference-hardware thesis.
A Crowded Inference Chip Race
Etched picked a high point of the market to step forward. Cerebras completed the year’s breakout IPO in May, raising $5.5 billion; Groq confirmed a $650 million raise on June 22. The cloud giants keep expanding their in-house silicon — Amazon, Google, and Microsoft each run their own AI chip programs — and OpenAI unveiled its first custom chip, built with Broadcom, on June 24.
In other words, “beyond Nvidia” has graduated from slogan to a real racetrack, and Etched is one of the few independent players on it with $1 billion in orders and silicon that has already been manufactured.
Why It Matters Now
For teams buying compute, Etched’s progress marks another step in diversifying inference supply: as the efficiency claims of specialized systems get validated, the unit cost of inference has room to fall. But the real test is just beginning — between customer testing and large-scale deployment sit yields, supply chains, and the realities of operating in live data centers. The $1 billion order book is a milestone, not a finish line. The next signal to watch: which customers are willing to move production inference workloads onto Frontier clusters.
Sources
- Nvidia competitor Etched hits $5B valuation, $1B in sales for AI chip — TechCrunch
- Etched official site
- Cerebras raises $5.5B in IPO — TechCrunch
- Groq confirms $650M raise — TechCrunch
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.