The chip startup that bet AI would outgrow general-purpose GPUs just put numbers on the table. Etched, an Nvidia competitor founded in 2022, came out of stealth Tuesday with a $5 billion valuation and $1 billion in booked contract orders, according to TechCrunch AI. The announcement follows TSMC successfully manufacturing Etched’s first chip earlier this year.
What stands out here is the speed of the turnaround. Two years ago, TechCrunch AI reports, the founders couldn’t get a single major investor to bite. Today they’re sitting on a nine-figure order book and a cap table full of AI royalty.
What Etched actually sells
Etched isn’t shipping bare chips. It’s selling full systems it calls “frontier inference clusters” – bundles that pair its custom silicon with purpose-built racks and software. The pitch: run frontier models faster, cheaper, and with better power efficiency than the competition.
The target is inference, and that choice matters.
- Inference is what happens after you hit enter on a prompt. It’s the model generating your answer.
- It’s now the biggest bottleneck and the biggest cost center for AI companies serving customers at scale.
- Anyone who can make inference cheaper and faster has a direct line to every lab’s budget.
That’s why investors are circling. Etched is currently testing that first product with customers, so the $1 billion is contracted demand, not delivered revenue yet.
The money and the names
Etched has raised $800 million total. The headline tranche was a previously unannounced $500 million round, closed in December at a $5 billion post-money valuation and led by Stripes.
The investor list reads like a who’s who:
- Firms: VentureTech Alliance, Jane Street, Hudson River Trading, Two Sigma, Ribbit Capital, and Stripes.
- AI heavyweights (angels): Andrej Karpathy, Geoffrey Hinton, Fei-Fei Li, Arthur Mensch, and Scott Wu.
- Billionaires: Stanley Druckenmiller and Peter Thiel.
Co-founders Gavin Uberti and Robert Wachen both dropped out of Harvard and became Thiel fellows to start the company. Worth noting: TechCrunch AI points out the “coming out of stealth” framing is generous, since the pair have been talking publicly about their chip plans since 2024, when Etched had already raised more than $125 million.
From nearly broke to bidding war
The backstory is the interesting part. On Patrick O’Shaughnessy’s “Invest Like the Best” podcast, the founders said that back in 2023 they wrote a 30-page memo arguing AI would eventually need specialized chips, not just general-purpose GPUs. Every major investor they pitched passed. The company was reportedly running month-to-month, close to out of cash.
The thesis they couldn’t sell in 2023 is now the hottest trade in tech.
Why this matters
This is significant because it confirms a shift that’s been building all year: the AI chip market is no longer a one-horse race, and inference is where the challengers are aiming. Nvidia still dominates training, but the money is waking up to the fact that serving models is a different problem with different economics.
The competitive picture backs that up:
- Cerebras delivered the first breakout AI chip IPO of the year.
- Groq just raised $650 million.
- Amazon, Google, and Microsoft all build their own in-house AI silicon.
- Even OpenAI announced its first custom chip, built by Broadcom.
Etched’s bet is narrower than most. Instead of chasing general-purpose flexibility, it’s building specialized hardware for one job and betting that specialization wins on cost and speed. If the customer tests go well, that $1 billion order book becomes a proof point that purpose-built inference silicon can peel real demand away from GPUs.
What to watch next
The order book is booked, not delivered. The real test is whether those first customers confirm the performance and efficiency claims once the clusters are running in production. If they do, expect more capital to flood toward inference-specific chips and more pressure on Nvidia’s margins in the part of the market that’s growing fastest.
For the full breakdown, check the original report at TechCrunch AI.