Chip startup Etched defies skeptics with $10.3B valuation
Chip startup Etched defies years of doubt after closing a $300 million Series C at a $10.3 billion valuation, TechCrunch reports. Led by Sequoia with Andreessen Horowitz, SK Hynix, Jane Street, and others, the Harvard-dropout-founded firm says its systems speed AI inference without relying on GPUs.
Key Takeaways
- Etched closed a $300 million Series C at a $10.3 billion valuation, roughly doubling from $5 billion in about seven months.
- Sequoia led the round; Andreessen Horowitz, SK Hynix, Jane Street, and Diffusion Capital also participated.
- The company says custom prefill chips and "cluster scale memory" accelerate inference on any AI model as full systems.
- Etched reports $1 billion in booked orders and early client testing after TSMC manufactured its silicon.
What did chip startup Etched announce, and why does it matter?
According to TechCrunch, co-founder and COO Robert Wachen confirmed the Series C and $10.3 billion valuation.
Etched was valued at $5 billion in December after raising $500 million. The company says this is the highest valuation ever for a Sequoia-led Series C.
Last month, Etched said it had successfully manufactured its chips, that clients were testing its first full systems, and that it had booked $1 billion in orders. For readers tracking AI hardware in our Future Tech & AI Wonders coverage, the raise underscores investor conviction that specialized inference gear can challenge GPU-centric stacks.
How does Etched claim to speed up inference without GPUs?
Etched designs systems around two inference stages: prefill and decode. Prefill understands the prompt and is compute-heavy. Decode generates output tokens and needs massive memory bandwidth.
Wachen says Etched built a prefill chip for "low-voltage inference," running at much lower voltage to cut heat and pack more transistors. For decode, it created new memory and interconnect technology it calls cluster scale memory, letting many chips share a fast, low-latency memory pool.
The firm argues its systems can run any AI model — including Mixture of Experts designs like DeepSeek and Qwen, plus non-transformer architectures such as Mamba — pushing back on the perception that its silicon only fits specific large language models.
Who is backing Etched, and what comes next?
Sequoia led the $300 million round. Andreessen Horowitz, SK Hynix, Jane Street, and Diffusion Capital joined, alongside earlier investors. Other backers named by TechCrunch include Peter Thiel, Andrej Karpathy, Dylan Field, and Amjad Masad.
Founded in 2022 by Harvard dropouts CEO Gavin Uberti, Wachen, and CTO Chris Zhu, Etched now has about 400 employees and operates a 2 megawatt data center. Access has mostly been limited to investors and early customers via private demos.
Wachen says mass-producing and delivering rack systems will still be hard. Even so, he notes that skeptics who tried the hardware — including AI researchers such as Andrej Karpathy, Noam Brown, and Geoffrey Hinton — came away excited about what they saw.