The capital injection arrives as Etched shifts from experimental design to commercial delivery. The startup recently confirmed it has successfully manufactured its first homegrown chips and secured $1 billion in pre-orders. By targeting the specific compute-intensive stages of AI inference—the prefill and decode phases—Etched claims its hardware can outperform traditional architectures by running at lower voltages and utilizing a proprietary cluster-scale memory system.
Etched hits $10.3 billion valuation as AI chip bets pay off
Seven months after securing a $5 billion valuation, AI hardware startup Etched has doubled its worth, closing a $300 million Series C round led by Sequoia. The company, founded by three Harvard dropouts, is now backed by a high-profile roster including Peter Thiel and Andrej Karpathy to scale its specialized inference chips.

While the industry initially questioned the viability of building chips tailored exclusively for transformer-based models, the founders have pushed back against the skepticism. COO Robert Wachen notes that their hardware is adaptable, supporting both Mixture of Experts models and non-transformer designs like Mamba. With 400 employees and a dedicated 2-megawatt data center, the firm is moving beyond its humble origins in a friend’s garage. Despite the momentum, the team acknowledges the significant engineering and logistical hurdles remaining to bring their rack systems to full-scale mass production.


Comments (0)
No comments yet. Be the first!