AI chip startup Etched defies skeptics, raises $10.3 billion valuation from major investors

engraved, The AI ​​chip startup, founded by three Harvard dropouts in 2022, has closed a $300 million Series C funding round at a $10.3 billion valuation, co-founder and COO Robert Washen told TechCrunch.

The round was led by Sequoia, and also participated in Andreessen Horowitz, SK Hynix, Jane Street and Diffusion Capital, along with other previous investors. Other backers of the company include names like Peter Thiel, Andrei Karpathy, Dylan Field, Amjad Massad, and others.

Etched was previously valued at $5 billion in December when it raised a $500 million round, meaning it has doubled its value in about seven months. The company says this is the highest valuation ever for a Sequoia-led C-series. Last month, Etched announced that it had acquired… It succeeded in manufacturing its own chips locallyits first complete systems have been tested by customers, and it has already booked $1 billion in orders.

The engraved was launched at a time when it was an idea Build a custom chip for AI models It is based on Transformer technology (the architecture behind most modern AI systems, including ChatGPT and Cloud). It is considered Wild if not foolish. The company is still struggling with the perception that its products — which are sold as complete systems, not just chips — include chips designed to run only specific LLM programs.

Fashion explains that this is not the case. The systems can run any AI model, including Mixture of Experts models like DeepSeek and Qwen — an architecture that divides tasks across specialized submodels rather than relying on one large model — as well as non-transformer designs like Mamba, built on a different model. Basic architecture Known as the state space model. (Interestingly, the idea of ​​etching parts of a particular AI model directly into silicon to boost performance is not considered so far-fetched anymore. Google is said to be pursuing the same concept through its Frozen V2 chip For Gemini.)

However, Etched’s claim to fame today is that it designed two new components from scratch to speed up inference, the computing process that occurs after a user submits a prompt.

“Inference is built in two stages, prepackaging and decoding,” says Washen. The “prefill phase” involves understanding the prompt, including the context. It’s a calculation and intensive process. The “Decoder” phase generates the output codes (the actual answer the user sees). It requires less computation but requires huge amounts of memory.

Etched has created a pre-packaging chip that runs “dramatically” faster, he promises, “by operating at a much lower voltage than any other AI chip. We call this low-voltage heuristics.” The lower voltage generates less heat, allowing the chip to pack more transistors.

For the decoding process, Etched created a new type of memory and “an interconnection technology that we call cluster memory. It allows many chips to communicate together and use a shared memory pool with very high speed and low latency,” he says. The result, Etched promises, is high speeds but lower costs.

Because the startup was launched before most of the tech world (besides Nvidia) understood the specialized computing needs of AI, the founders, CEO Gavin Uberti, Wachen, and Chris Zhu, He faced a lot of skeptics Who remained skeptical even after the company announced that the first batch of silicon had been successfully manufactured by TSMC.

Much of this comes from the small number of people who have access to the systems. Until now, access has been limited to early investors and customers. In fact, that’s how Etched got on its list of famous investors in the first place — by showing them private demos in its office.

“Andrei Karpathy from Anthropic, Noam Brown from OpenAI, Geoffrey Hinton, as well as all the investors in the funding round — these are all people who have already tried the hardware and are very excited about it,” says Washen.

However, it has been a long and difficult road, and there is still a long way to go until the shelving systems are mass-produced and delivered. The famous trio He left Harvard University To launch Etched, without knowing how to raise money (let alone how much money they would need) or how to hire.

“We had no idea how hard it would be,” Fashion says. “I think we still have to be humble about what it will take to get to really broad scale.”

Washen recalls arriving in the Bay Area after telling his parents he was leaving school to start a startup, without an office or apartment arranged. He was sleeping on the floor of a friend’s unfurnished house.

“I remember staying at my friend’s house that they were about to sell, using a towel as a blanket,” he laughs. Eventually, the founders set up the servers they needed to run their chip design tools in an early employee’s garage, and “every time he needed to reboot, he would call his wife, and she would go over and hit the reboot button.”

Today, there are 400 people crowded into an office, and Etched runs a 2-megawatt data center there. It just opened a new 80,000-square-foot, 10-megawatt facility in Milpitas, a short distance from its main office in San Jose.

“Today we run tokens in our lab, working with some of the largest AI companies in the world,” he says.

Fashion now also has a blanket, bedding, “and even a pillow. Multiple pillows,” he jokes.

Most importantly, he and his co-founders never let doubters stop them. “We’ve come a long way. It’s a very different world. But I think when you really believe something is possible, and you work at it for a long time, you can do it.”

Note: This story has been updated to include a reference to the new 10-megawatt facility.