Etched Raises $700 Million as Valuation Jumps to $21 Billion
AI-chip startup Etched has raised $700 million at a $21 billion valuation, more than doubling its valuation in less than a month as investors pour capital into companies trying to challenge Nvidia in the fast-growing market for artificial-intelligence inference hardware.
The latest round was led by quantitative trading firm Jane Street, with participation from Kleiner Perkins, Sequoia Capital, Andreessen Horowitz, Tiger Global, Bain Capital Ventures, Neo and Blackstone.
Etched was valued at about $10.3 billion in a $300 million funding round announced in July. Its new $21 billion valuation therefore represents an unusually rapid jump for a semiconductor company whose first commercial hardware is only now beginning to reach customers.
Founded in 2022 and based in San Jose, California, Etched is developing specialised hardware for AI inference — the process through which trained AI models generate responses when users interact with them.
Rather than building a general-purpose graphics processor similar to Nvidia’s GPUs, Etched is designing its systems specifically around the computational demands of modern AI models. The company says that tighter specialisation can deliver higher throughput and greater power efficiency for increasingly large inference workloads.
Etched says its first silicon has returned from TSMC’s N4P manufacturing process and is now being validated inside rack-scale systems with customers. The company also says it has secured more than $1 billion in customer demand. Those figures are company-reported and have not been independently audited.
Jane Street is particularly important to the story because it is not only leading the funding round.
The trading firm has also become Etched’s first customer and has begun deploying the startup’s hardware, according to reporting from The Wall Street Journal and Reuters. That gives the investment a commercial relationship beyond a conventional venture-capital bet.
Etched has grown to more than 400 employees, recruiting engineers from companies including Nvidia, Google’s TPU organisation, Broadcom, SK Hynix and TSMC.
The company is entering an increasingly competitive part of the AI infrastructure market.
Nvidia remains overwhelmingly dominant in AI computing, but companies including Etched, Cerebras and other specialised-chip developers are betting that the enormous cost and electricity requirements of running AI models will create demand for hardware designed specifically for inference.
For customers deploying increasingly large AI systems, the competition is moving beyond raw computing performance toward cost per generated token, energy efficiency, latency and the ability to run extremely large models at scale.
Etched’s sudden rise also highlights how aggressively investors are funding that challenge.
A company valued at just over $10 billion in July is now being valued at $21 billion in August, before its technology has been tested at anything approaching Nvidia’s commercial scale.
The next test will be whether Etched can turn its early customer commitments and technically promising hardware into a sustainable semiconductor business.
