- Etched raised $700 million, bringing its value to $21 billion and thus doubling in less than a month.
- Jane Street not only led the funding round but also served as Etched’s first production customer by operating its own rack within the company.
- Etched has received orders totaling over $1 billion, and two of its main competitors have lost their independence.
Etched, a US-based startup that makes processors built to run trained AI models, has raised a $700 million funding round led by Jane Street, doubling its valuation to $21 billion within a month.
Sequoia Capital, Andreessen Horowitz, Tiger Global, Peter Thiel, Bain Capital Ventures, Neo, Stripes, Primary, Positive Sum, Diffusion, Argo, and Blackstone all acted as investors once again. Kleiner Perkins became a new investor.
In a blog post announcing the round, Jane Street said it had tested the chip and was pleased with the early results. The method that Etched uses for inference provides the precision required to support our most demanding workloads.
From rejection to record-breaking
The startup was founded in 2022 by Gavin Uberti, Chris Zhu, and Rob Wachen following Uberti’s departure from Harvard. In 2023, the founders had difficulty getting investors to answer the phone, and the company was nearly out of cash. Looking back at that situation three years later, it seems hard to believe.
Etched, now based in San Jose and with more than 400 employees, has raised almost $2 billion to date. In July, it achieved a valuation of $10.3 billion, a figure which Sequoia described as representing its largest Series C to date, and since then, the valuation has roughly doubled.
Etched produces processors designed to run already-trained AI models rather than train them. The reasoning is simple: inference will one day require more computing power than training.
Wachen describes the process in two stages: the first is a compute-intensive prefill phase in which the system reads and understands the prompt, and the second is a memory-intensive decode phase during which the answer is generated.
The prefill chip used by Etched requires less than half the voltage of ordinary AI accelerators, which allows it to include more transistors without overheating. As a result, trading firms that use latency-sensitive models can obtain more responses per rack without adding extra cooling.
Competition in the market has decreased
Since July, Etched’s position has become stronger, which can be attributed to external events. In December 2025, Nvidia entered into a deal, stated to be worth tens of billions, under which it licensed Groq’s inference technology, and Jonathan Ross, Groq’s founder, and most of the engineering team also joined Nvidia.
In May 2026, Cerebras had its initial public offering at a valuation of nearly $56 billion, in exchange for giving up its status as an independent company and accepting greater oversight. Nvidia remains the market leader, accounting for around 80% of the demand for AI training chips.
Over the past six months, many venture-backed companies have either been acquired or gone public, leaving fewer independent companies.
Etched is one of the very few independent companies still in business and offers complete frontier inference clusters rather than just individual chips; its approach is therefore more akin to what Nvidia refers to as AI factories than to that of traditional component suppliers.
Etched still has to demonstrate the results from its order pipeline
Etched has already begun shipping its chips and has over $1 billion in orders, but it has not announced any revenue. In 2026, the global market for AI inference chips was worth $20.5 billion and is expected to reach $37 billion by 2030.
The market value of Etched is based on its contracts, despite its limited financial history. Even though the company has a valuation of $21 billion, this gives it only a limited amount of time.
Etched will have to convert those contracts into working clusters for customers outside its investor group.
