Etched AI raises $300M Series C from Sequoia/a16z at $10.3B post-money valuation for inference-specific silicon.

AI chip startup Etched has announced the completion of a $300 million Series C funding round, achieving a post-money valuation of $10.3 billion. Sequoia Capital led the round, with participation from a16z, SK Hynix, Jane Street, and others.

The company's valuation has doubled in just seven months, following a $500 million funding round completed last December at a $5 billion valuation. Etched's first chip has been successfully manufactured by TSMC and has already secured $1 billion in orders. The company currently employs 400 people and operates a 2-megawatt data center.

This funding underscores growing capital investment in AI inference-specific hardware. The fresh capital will be directed toward mass production and customer deployment, benefiting Etched and its supply chain partners, while general GPU manufacturers face increasing competitive pressure in the inference segment.

Founded in 2022 by three Harvard dropouts, Etched has progressed rapidly from prototype to production scale. Co-founder and COO Robert Wachen began the venture working from a friend's floor. The company is advancing chip production in partnership with TSMC, with a specific focus on AI inference rather than training workloads.

Etched's capital strategy emphasizes a complete system solution rather than a single isolated chip. The approach optimizes both pre-filled, compute-intensive stages and decoding memory-intensive stages, reducing latency and costs through low-voltage, high-density chips and clustered shared memory interconnects. The goal is to capture inference workload market share by addressing the power consumption and cost bottlenecks inherent to general-purpose GPUs.

The competitive landscape includes dedicated inference chips from companies like Groq and Cerebras, as well as NVIDIA's dominant CUDA ecosystem. Etched is positioned in the critical expansion phase from prototype to mass production delivery, and has received personal endorsements from individuals at both Anthropic and OpenAI.

The emergence of specialized inference hardware is fundamentally driven by explosive demand for inference capabilities. A dedicated full-system solution can bypass the power and cost constraints that limit general GPUs in this segment.