Etched AI Soars to $10.3B, Silencing Skeptics with Major Investor Backing.

Etched, the AI chip startup founded in 2022 by three Harvard dropouts, has successfully concluded a Series C funding round, raising $300 million and achieving a valuation of $10.3 billion. This significant development was confirmed by co-founder and COO Robert Wachen in an exclusive statement to TechCrunch.

The latest investment round was spearheaded by Sequoia, with notable participation from Andreessen Horowitz, SK Hynix, Jane Street, and Diffusion Capital, alongside several earlier investors. The company's impressive roster of individual backers also includes influential figures such as Peter Thiel, Andrej Karpathy, Dylan Field, and Amjad Masad, among others.

This new valuation represents a remarkable doubling of Etched's worth in approximately seven months, following its $5 billion valuation in December when it secured a $500 million funding round. The company proudly states that this marks the highest valuation ever recorded for a Sequoia-led Series C. Just last month, Etched had announced significant milestones: the successful manufacturing of its proprietary chips, the commencement of client testing for its inaugural full systems, and the accumulation of $1 billion in pre-orders.

Etched emerged at a time when the concept of developing chips specifically optimized for AI models based on transformer technology—the foundational architecture for most contemporary AI systems like ChatGPT and Claude—was largely regarded as unconventional, if not entirely eccentric. The company continues to address the misconception that its products, which are sold as comprehensive systems rather than standalone chips, are designed exclusively to run specific large language models (LLMs).

Wachen clarifies that this perception is inaccurate. He explains that Etched's systems are versatile enough to execute any AI model, including sophisticated Mixture of Experts (MoE) architectures like DeepSeek and Qwen, which distribute tasks across specialized sub-models instead of relying on a single large one. Furthermore, they support non-transformer designs such as Mamba, built on a distinct underlying state-space model architecture. Interestingly, the notion of directly embedding parts of an AI model into silicon for enhanced performance is no longer considered far-fetched, with Google reportedly exploring a similar concept for its Gemini AI with the 'Frozen v2' chip.

At its core, Etched's current distinction lies in its ground-up design of two innovative components engineered to dramatically accelerate AI inference—the computational process that occurs after a user submits a prompt.

“Inference is built in two stages,” Wachen details, identifying them as “prefill and decode.” The "prefill phase" is dedicated to understanding the user's prompt, including its contextual elements, making it both mathematically and computationally intensive. Conversely, the "decode" phase is responsible for generating the output tokens, or the actual response seen by the user. While requiring less raw computation, this stage demands substantial memory resources.

Etched's solution for the prefill phase is a chip that operates "dramatically" faster, Wachen asserts, by utilizing "a much lower voltage than any other AI chip." This innovation, dubbed "low-voltage inference," significantly reduces heat generation, thereby enabling the chip to accommodate a greater density of transistors.

For the decode process, Etched has pioneered a novel memory and interconnect technology, which they term "cluster scale memory." Wachen explains that this allows numerous chips to seamlessly connect and access a shared memory pool with "very, very fast, low latency." The ultimate promise from Etched is a combination of exceptional speeds and reduced operational costs.

Given that the startup launched before the broader tech industry, with the notable exception of Nvidia, fully grasped the specialized computing demands of AI, the founders—CEO Gavin Uberti, Robert Wachen, and CTO Chris Zhu—encountered considerable skepticism. This doubt persisted even after the company announced the successful manufacturing of its initial silicon batch by TSMC.

Much of this skepticism stemmed from the limited access to Etched's systems, which initially was restricted to investors and early customers. Ironically, it was precisely through these private demonstrations, conducted in their office, that Etched managed to attract its distinguished list of investors.

“Andrej Karpathy from Anthropic, Noam Brown from OpenAI, Geoffrey Hinton, as well as all the investors in the funding round — these are all people who actually tried the hardware and are very excited about it,” Wachen proudly states, highlighting the endorsement from leading AI experts.

Despite these successes, the journey has been long and arduous, with further challenges ahead before the rack systems can be mass-produced and delivered. The trio famously decided to drop out of Harvard to establish Etched, initially lacking knowledge about securing capital—let alone the substantial funds they would require—or the intricacies of hiring a team.

“We had no idea how hard it was going to be,” Wachen reflected. “I think we still have to be humbled by what it will take to actually get to scale.”

Wachen vividly recalls his arrival in the Bay Area after informing his parents of his decision to leave academia for a startup, without having secured an office or even an apartment. He resorted to sleeping on the floor of an unfurnished friend's house.

“I remember staying in my friend’s house that they were about to sell, using a towel as a blanket,” he recounts with a laugh. The founders eventually set up the essential servers for their chip-design tools in the garage of an early employee, where “every time it needed to be rebooted, he would call his wife, and she would go and hit the reboot button.”

Today, the contrast is stark: 400 employees are actively working in a bustling office, and Etched operates a robust 2-megawatt data center. “We’re running tokens in our in our lab today, working with some of the largest AI companies in the world,” Wachen affirms, showcasing their current operational scale and partnerships.

Wachen lightheartedly adds that he now possesses a proper blanket, a mattress, and "a pillow even. Multiple pillows," joking about the significant improvement in his personal circumstances.

More significantly, he emphasizes that he and his co-founders never allowed the initial skepticism to deter them. “It’s come a long way. It’s a very, very different world. But I think, when you really think something’s possible, and you just work at it for a long time, you can do it.”