The Short Version: A Startup That Bet Big on One AI Model Type
Three Harvard dropouts started Etched in 2022 with a pretty specific bet. They believed the AI boom would run on something called transformer models, which form the foundation of many advanced AI systems such as ChatGPT and Claude. Most chip companies build general-purpose processors that can handle anything. Etched went the other direction: it builds custom chips tailored to transformer-based AI models, focusing on the part of AI that happens after you type a question.
The company calls that part "inference." When you submit a prompt to an AI model, the model does a bunch of computation to figure out what to say back. That computation happens in two stages, according to COO and co-founder Robert Wachen: "prefill and decode." Etched's chips are built to speed up both steps.
Big-name investors bought into the idea. Sequoia Capital led the funding round. Other participants in the funding were Diffusion Capital, Andreessen Horowitz, SK Hynix, and Jane Street.
Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter
Individual investors include Peter Thiel, Andrej Karpathy, Dylan Field, and Amjad Masad. The chips themselves are made by TSMC.
The three co-founders - Gavin Uberti, Chris Zhu, and Robert Wachen - left Harvard in 2022 to start the company. Their conviction has paid off so far.
A Valuation That Doubled in Under a Year
The new funding round values Etched at $10.3 billion. That is more than double the $5 billion valuation it carried just seven months earlier, when it raised $500 million in December.
One month before this article, Etched had already booked $1 billion in orders for its hardware. The company now employs 400 people and operates out of a main office with a data center capable of 2 megawatts of power.
Wachen is realistic about what comes next. He told TechCrunch that the company "had no idea how hard it was going to be" and that they still need to be "humbled by what it will take to actually get to scale." He also noted that those who have tested the hardware firsthand - among them Anthropic's Andrej Karpathy, OpenAI's Noam Brown, and Geoffrey Hinton - "are very excited about it."
Betting entirely on a single model architecture carries significant risk. If a new AI breakthrough shifts the industry away from transformers, Etched's specialized chips could lose relevance. However, the company's rapid growth and backing from top investors and AI pioneers suggest confidence that transformers will remain dominant for the foreseeable future. With $1 billion in orders already on the books, Etched is racing to scale production and deliver its chips to customers.
Etched's strategy also depends on its manufacturing partnership with TSMC, which produces the custom silicon.
Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets
