AI Chip Startup Etched Hits $10.3 Billion Valuation in Sequoia-Led Series C
Etched raised $300 million in a Sequoia-led Series C, doubling its valuation to $10.3 billion in seven months as it builds specialized hardware to compete in AI inference compute.
AI chip startup Etched has closed a $300 million Series C round led by Sequoia, valuing the company at $10.3 billion. The raise comes just seven months after the San Jose-based startup secured $500 million at a $5 billion valuation. Andreessen Horowitz, SK Hynix, Jane Street, and Diffusion Capital also participated, alongside angel investors including Peter Thiel, Andrej Karpathy, and Dylan Field, according to TechCrunch.
Etched was founded in 2022 by Harvard dropouts Gavin Uberti, Robert Wachen, and Chris Zhu. The company builds hardware designed specifically for AI inference, the process of running a trained model to respond to user queries. Unlike GPUs, which handle both training and general-purpose computing, Etched's chips and integrated server systems are purpose-built to accelerate model outputs. The company manufactures its silicon through TSMC and says it has already secured $1 billion in pre-orders.
The company tackles inference performance through two distinct hardware components. For the prefill stage, which processes input prompts and context, Etched designed a custom chip that runs at reduced voltage to limit heat and increase transistor density. For the decode stage, which generates output text, the company developed what it calls cluster-scale memory, enabling multiple chips to share memory pools with lower latency. Co-founder Robert Wachen noted that the systems support multiple model architectures, including Mixture of Experts models like DeepSeek and Qwen, as well as state-space models like Mamba.
Etched now employs 400 people and recently opened an 80,000 square foot, 10-megawatt testing facility in Milpitas, California. Access to full hardware systems remains limited to early partners and investors, with initial units currently in client testing.
The Cognarah Angle
The rapid ascent of inference-specialized hardware startups like Etched reflects a real and growing tension in the AI infrastructure stack: generic GPU clusters are becoming an expensive bottleneck. For African tech hubs, where energy availability is constrained and cloud costs arrive in hard currency, inference speed and power efficiency are not abstract benchmarks. They determine whether local developers can build commercially viable services on frontier models, or remain priced out by compute bills payable in dollars to data centers thousands of kilometers away.
The overwhelming majority of African startups depend on API calls to foreign cloud providers, absorbing premium inference costs routed through North American or European infrastructure. If purpose-built chips genuinely deliver the cost and efficiency reductions their backers claim, cloud providers and regional data centers could eventually pass some of those gains down the stack. But a high valuation in Silicon Valley does not equal hardware accessibility in Nairobi, Lagos, or Accra. As long as specialized chips remain concentrated inside Western hyperscalers serving tier-one technology companies, African developers will keep paying legacy rates for compute designed and priced elsewhere.
That is the uncomfortable question this funding round raises for the continent. Capital flooding into Silicon Valley chip designers produces headlines, not lower cloud invoices. African founders need the efficiency gains to show up in actual billing, not in TechCrunch valuations. So far, there is no sign that is part of the plan.
Reporting sourced from TechCrunch. Analysis and Cognarah Angle are Cognarah's own.
Written by
StacyAI-assisted news curation. Every story is reviewed by our editors before publication.
