According to Watchful AI monitoring, Etched, an AI chip startup founded by three Harvard dropouts in 2022, announced the completion of a $300 million Series C financing round, with a post-money valuation of $10.3 billion. The round was led by Sequoia Capital, with participation from a16z, SK Hynix, Jane Street, and Diffusion Capital, among others. Individual investors include Peter Thiel, Andrej Karpathy, and Dylan Field, among others.

The company had just raised $500 million at a $5 billion valuation in December last year, doubling its valuation in seven months. Sequoia called it the highest-valued Series C round they have led in their history. Last month, Etched announced that its first chip had been successfully manufactured by TSMC, with the initial complete systems currently undergoing customer testing, locking in $1 billion in orders. The company currently has 400 employees and a 2-megawatt data center, and is collaborating with several leading AI companies globally for testing. Co-founder and COO Wachen jokingly remarked that compared to the early days of entrepreneurship when they used towels as blankets on a friend's floor, they now not only have a mattress but even several pillows.

Etched's core selling point is a full-system solution designed specifically for AI inference—not just selling chips. Wachen explained that inference is divided into two stages: "pre-filling" and "decoding," with the pre-filling stage being math and computation-intensive, while the decoding stage requires massive memory. Etched has thus developed dedicated pre-filling chips (operating at significantly lower voltages than competitors to achieve higher transistor density) and in-house cluster-level shared memory interconnect technology (enabling multiple chips to access a unified pool of memory at ultra-low latency), aiming to significantly reduce costs while reducing latency.

Facing market skepticism about "chips only being able to run specific models," Wachen responded that the system can run any AI model, including hybrid expert models like DeepSeek and Qwen, as well as non-Transformer architecture models like Mamba. The company has been gradually building its reputation through in-person demonstrations with investors and early clients—Andrej Karpathy from Anthropic, Noam Brown from OpenAI, and Geoffrey Hinton all highly praised the hardware after personal testing. While the fully-equipped rack systems are still some distance away from large-scale production and delivery, Wachen said, "When you truly believe something is possible and work hard towards it for a long time, you can achieve it."

Click the original article link below to join the Insight X Beating · Feishu AI News Channel, which monitors global AI trends and news 24/7.