Etched Is Fielding Funding Offers at $40 Billion to $50 Billion — Roughly Double Its $21 Billion Round From Weeks Ago

Etched, the startup building custom silicon for AI inference, is reviewing unsolicited investment offers that would value the company at \$40 billion to \$50 billion, according to people familiar with the discussions cited by TechCrunch. That is roughly double the \$21 billion valuation Etched set just weeks ago, when Jane Street led a \$700 million round.

The talks are preliminary, no new financing has been announced and terms could change; Etched declined to comment. The offers are not evenly distributed, either. Top-tier investors are said to be bidding around \$40 billion, while lesser-known backers have pushed toward \$50 billion — a split that says a great deal about who is underwriting the AI hardware boom and on what terms.

The revaluation has been fast even by current standards. In July, Sequoia led a \$300 million Series C at a \$10.3 billion valuation. Weeks later, Jane Street led a \$700 million round at \$21 billion. Bids in the \$40-50 billion range would represent roughly a fourfold step-up in about three months, for a four-year-old company with no production revenue on the books.

What investors are actually buying is a backlog and a claim. Etched said in July that it had secured \$1 billion in customer orders after fabricating a test chip at TSMC, and Jane Street — both an investor and an early customer — has taken delivery of an early system. The company says its chips can process more tokens faster and at lower cost than Nvidia's, a performance assertion that has not been independently verified. For a quantitative trading firm like Jane Street, shaving inference latency translates directly into trading advantage, which is why its dual role as backer and buyer matters more than a typical strategic check.

The physical footprint is modest for the valuation. Etched operates a 10-megawatt data center in Silicon Valley and maintains a facility in Taiwan to sit close to manufacturing partner TSMC. It employs roughly 400 people, about 15% of whom previously worked at Nvidia, according to The Wall Street Journal. Founders Gavin Uberti and Chris Zhu met in an advanced math course at Harvard; COO Robert Wachen was Uberti's roommate. All three dropped out to build the company.

The runway math explains why Etched would raise again so soon. Building complete systems around proprietary chips — rather than selling standalone silicon — is capital-intensive, and a person familiar with the offers noted that a round similar in size to the last one could fund as much as three and a half years of operation.

There are two ways to read a bid like this. The generous one: inference is where AI economics actually settle, a credible Nvidia challenger with a \$1 billion order book is genuinely scarce, and pricing on backlog rather than revenue is what early infrastructure always looks like. The less generous one: a valuation that doubles in eight weeks on a test chip is the frothy end of a very long boom, and the spread between \$40 billion and \$50 billion — established firms versus unknown backers — is itself a signal about conviction. If Etched closes near the top, the whole inference-ASIC comp set re-rates. If it walks away and the bids cool back toward \$21 billion, that is the first real markdown in the category.