AI chip startup Etched raises $300 million in its Series C round at a $10.3 billion valuation

ME AI News: According to monitoring by Beating, Etched, an AI chip startup founded in 2022 by three Harvard dropouts, has announced the completion of a $300 million Series C round, bringing its post-money valuation to $10.3 billion. Sequoia Capital led the round, with participation from a16z, SK Hynix, Jane Street, Diffusion Capital, and individual investors including Peter Thiel, Andrej Karpathy, and Dylan Field. The company previously raised $500 million in December last year at a $5 billion valuation—doubling its valuation in just seven months. Sequoia called this the highest-valued Series C round it has ever led. Last month, Etched announced that its first chip had been successfully manufactured by TSMC, with the first complete systems currently undergoing customer testing and $1 billion in orders already secured. The company now employs 400 people and operates a 2-megawatt data center, while collaborating with leading AI companies worldwide on testing. Co-founder and COO Wachen joked that compared to the early days when they slept on friends’ floors using towels as blankets, “now we not only have mattresses, but several pillows.” Etched’s core offering is a full-system solution designed specifically for AI inference—not just chips. Wachen explained that inference consists of two phases: “prefill,” which is mathematically and computationally intensive, and “decoding,” which demands massive memory. To address this, Etched developed dedicated prefill chips that achieve higher transistor density at significantly lower voltages than competitors, along with proprietary cluster-level shared memory interconnect technology that enables multiple chips to access a unified memory pool with ultra-low latency, aiming to drastically reduce both cost and latency. In response to market skepticism that chips can only run specific models, Wachen noted the system supports any AI model—including mixture-of-experts models like DeepSeek and Qwen, and non-Transformer architectures like Mamba. The company has been building its reputation through in-person demos with investors and early customers; Andrej Karpathy from Anthropic, Noam Brown from OpenAI, and Geoffrey Hinton all expressed strong approval after hands-on hardware trials. While fully configured rack systems are still some distance from mass production and delivery, Wachen said, “When you truly believe something is possible and work at it for a long time, you can make it happen.” Click the link below to join Beating’s Feishu AI News Channel for 24/7 real-time monitoring of global AI trends and news. (Source: BlockBeats)