Etched just doubled its valuation in about seven months. The AI chip startup, founded by three Harvard dropouts in 2022, has closed a $300 million Series C at a $10.3 billion valuation, co-founder and COO Robert Wachen told TechCrunch AI. Sequoia led the round, and the company says it’s the highest valuation ever for a Sequoia-led Series C.
That’s a fast climb. Etched was valued at $5 billion back in December when it raised $500 million. According to TechCrunch AI, this new round pulled in Andreessen Horowitz, SK Hynix, Jane Street, and Diffusion Capital, plus earlier backers. The investor list also reads like a who’s who of AI: Peter Thiel, Andrej Karpathy, Dylan Field, and Replit’s Amjad Masad, among others.
What Etched actually builds
When the company launched, the idea of designing a chip specifically for transformer-based AI models (the architecture behind ChatGPT and Claude) was seen as reckless. The bet looks a lot smarter now. Google is reportedly chasing a similar concept with its Frozen v2 chip for Gemini.
Etched still fights one persistent misconception: that its hardware only runs specific large language models. Wachen says that’s wrong. The systems, sold as full racks rather than bare chips, can run any AI model. That includes Mixture of Experts designs like DeepSeek and Qwen, which split work across specialized sub-models, and non-transformer architectures like Mamba.
The company’s real pitch is speed on inference, the computation that happens after you hit enter on a prompt. Wachen breaks it into two stages:
- Prefill: the chip reads and understands your prompt and its context. It’s heavy on raw compute.
- Decode: the chip generates the answer you actually see. Lighter on compute, but hungry for memory.
Etched built a separate chip for each. Its prefill chip runs at much lower voltage than rival AI chips, which Wachen calls “low-voltage inference.” Less voltage means less heat, which means more transistors packed onto the chip. For decode, the company created what it calls “cluster-scale memory,” an interconnect that lets many chips share one fast, low-latency memory pool. The promise: higher speeds at lower cost.
Why this matters
Nvidia owns AI compute right now, and almost every serious challenger is trying to chip away at inference costs rather than training. Etched is one of the few that got real silicon out the door. Last month it said TSMC had successfully manufactured its chips, that clients were testing full systems, and that it had already booked $1 billion in orders. Bookings like that are what turn a hardware story from a pitch into a business.
What stands out here is who vouched for the product. Much of the skepticism around Etched came from how few people had touched the hardware, since access stayed limited to investors and early customers. That’s also how the company won its famous backers, through private demos in its office. Wachen name-checks Karpathy, OpenAI’s Noam Brown, and Geoffrey Hinton as people who tried the systems and walked away impressed.
The road behind and ahead
The founders, CEO Gavin Uberti, Wachen, and Chris Zhu, took a rough path. Wachen recalls landing in the Bay Area with no office and no apartment, sleeping on a friend’s floor with a towel for a blanket. The team ran chip-design tools on servers stuffed in an early employee’s garage, and “every time it needed to be rebooted, he would call his wife, and she would go and hit the reboot button.”
Today Etched runs 400 people, a 2 megawatt data center at its San Jose office, and a new 80,000 square-foot, 10MW facility in nearby Milpitas. “We’re running tokens in our lab today, working with some of the largest AI companies in the world,” Wachen said.
He’s realistic about what’s left. Mass-producing and shipping rack systems is a different challenge than proving a chip works. “We had no idea how hard it was going to be,” he said. “I think we still have to be humbled by what it will take to actually get to scale.”
Watch the delivery timeline. A $1 billion order book means nothing until those racks land in customer data centers. If Etched hits that, the challenger-to-Nvidia narrative gets a lot more real. Full details are in the original TechCrunch AI report.