How three Harvard dropouts built a $21B AI chip challenger to Nvidia

The trio founded Etched in 2022 and initially struggled to convince investors that three young founders with little semiconductor experience could build sophisticated AI hardware.

Among technology investors, the prevailing wisdom had been "don't back the kids in chips," Sequoia Capital general partner Sonya Huang told The Wall Street Journal. "There was a sense that these kids think they can do anything, and maybe on the internet, they can, but that in chips, startups are guilty until proven innocent."

Four years later, the San Jose, California-based startup has more than 400 employees, has booked more than $1 billion in orders and has begun shipping its AI systems. Its latest breakthrough came in August, when quantitative trading firm Jane Street led a $700 million funding round that valued Etched at $21 billion, more than double its $10.3 billion valuation less than a month earlier, according to Reuters. Jane Street is also its first customer.

"They move faster than anyone I’ve ever dealt with, especially in hardware, which typically moves really slowly," Etched board member Patrick O’Shaughnessy said.

Etched co-founders from left Robert Wachen, Gavin Uberti, and Chris Zhu. Photo courtesy of Etched

From Harvard classrooms to a chip startup

Etched's story began at Harvard, where Uberti and Zhu met in Math 55, a notoriously difficult undergraduate course in theoretical mathematics.

Wachen, a coder who lived on the same floor as Uberti, was running an education technology company. He later ran Prod, a student-run nonprofit startup accelerator in Cambridge, Massachusetts, that produced companies including the world's youngest self-made billionaires' AI hiring platform Mercor, before co-founding Etched.

The three eventually dropped out of Harvard to pursue the startup.

The founders had little experience in the industry, and investors and potential customers questioned whether they could succeed. Wachen recalled arriving in the San Francisco Bay Area after telling his parents he was leaving Harvard to start a company, without an office or apartment arranged. "I remember staying in my friend's house that they were about to sell, using a towel as a blanket," he told tech site TechCrunch.

The founders initially kept servers needed for chip-design work in an early employee's garage. When one needed restarting, the employee would call his wife and ask her to manually reboot it. "We had no idea how hard it was going to be," Wachen said. "I think we still have to be humbled by what it will take to actually get to scale."

Betting on AI inference

Etched was founded the same year OpenAI released ChatGPT, which helped fuel a surge in demand for the computing power needed to develop and operate generative AI. Nvidia emerged as the dominant supplier of processors powering the boom, while Etched made a narrower bet on inference.

Training is how an AI model learns. Inference happens afterward, every time someone asks a trained model a question and it generates an answer. As AI services attract more users, that process happens constantly, making the speed, energy consumption and cost of inference increasingly important.

Etched set out to build hardware specifically designed to make inference faster and more efficient. Its technology aims to let processors perform intensive calculations using less electricity while allowing multiple chips to share information more quickly.

After receiving its test chips from Taiwan Semiconductor Manufacturing Company, the world's largest contract chipmaker, Etched said it took just 44 days to get them running AI workloads, compared with a process that typically takes six months or longer.

The company says some communications operations that take about 4,000 nanoseconds on Nvidia's Blackwell processors can be completed in about 700 nanoseconds by its hardware. "We may be the only AI chip that was built by a startup that was successful on the first try," Wachen said.

By late 2025, investor skepticism was giving way to increasingly large bets on the company. Etched raised $500 million in December at a $5 billion valuation. By June this year, it said it had successfully produced its first chips and secured more than $1 billion in orders. A $300 million Series C round led by Sequoia in July pushed its valuation to $10.3 billion. That same month, Jane Street became Etched's first customer after testing its hardware and receiving its first server rack, before leading the latest funding round that helped more than quadruple the startup's valuation in eight months.

Etched now operates a two-megawatt data center at its San Jose office and has opened an 80,000-square-foot, 10-megawatt facility in nearby Milpitas. Around 15% of Etched's roughly 400 employees previously worked at Nvidia, while its broader team includes engineers from Google's TPU program, Broadcom, SK Hynix and TSMC.

Systems engineer Anthony Gamerman joined Etched in April after watching several people he admired leave Nvidia for the startup. "One leaves, and then another one, you hear a couple weeks later, and you go, 'Whoa, I can't believe so-and-so is heading over to Etched,'" he said.

Nvidia CEO Jensen Huang speaks during a doorstep after attending the AI Ecosystem Reception in Tokyo on July 16, 2026. Photo by Reuters

A challenger to Nvidia

Etched's rapid rise has positioned it among a growing group of companies trying to loosen Nvidia's grip on AI computing, but the gap remains enormous.

Nvidia, world's second most valuable company behind Apple, Nvidia has more than 80% of the market share for AI chips, according to CNBC's estimates. Its advantage extends beyond its processors to its widely used CUDA software ecosystem, while its hardware can handle both AI training and inference.

Etched is focused primarily on inference and is only beginning large-scale commercial deployment. even though it has claimed its chips can run some AI workloads more than 10 times faster than Nvidia's GPUs, its performance claims have yet to undergo extensive independent benchmarking.

"Etched has impressive technology and serious customer interest, but semiconductor history is littered with brilliant chips that never became great businesses," Michael Ashley Schulman, partner at Cerity Partners, told Reuters.

For the founders, however, the contrast with Etched's early days is already stark.

"We’re running tokens in our lab today, working with some of the largest AI companies in the world," Wachen said, quipping that he now has a blanket, a mattress "and a pillow even. Multiple pillows."

"It's come a long way. It's a very, very different world," he said. "But I think, when you really think something's possible, and you just work at it for a long time, you can do it."