A sudden jump in valuation
Etched said it has raised $700 million at a $21 billion valuation, with Jane Street leading the round after testing and buying the startup’s AI hardware. The quant trading firm has also installed Etched’s first shipped AI cluster system in its own data center.
The speed of the repricing is striking even by current AI market standards. Etched was valued at $5 billion in December, then raised a $300 million Series C at a $10.3 billion valuation in July. Roughly a month later, investors have doubled that figure to $21 billion, an increase of nearly $11 billion.
Other backers named by the company include Kleiner Perkins, Sequoia Capital, Andreessen Horowitz, Peter Thiel, Tiger Global, Bain Capital Ventures, Neo, Stripes, Primary, Positive Sum, Diffusion, Argo, and Blackstone.
The key figures are simple:
- New round: $700 million
- New valuation: $21 billion
- July Series C: $300 million at $10.3 billion
- December valuation: $5 billion
- First shipped cluster: installed by Jane Street
Why inference hardware is the focus
Etched sells full systems rather than only individual chips. It calls them “frontier inference clusters,” a phrase that refers to infrastructure built to run advanced AI models after a user sends a prompt. Inference is the process of producing an answer from a trained model; it is different from training, which is the earlier stage in which the model learns from data.
Etched co-founder and COO Robert Wachen described inference as having two parts. The first is the “prefill” stage, where the system reads and understands the prompt and its context. This is heavy in mathematical computation. The second is the “decode” stage, where the model generates output tokens, or the pieces of text that make up the response the user sees. Decode depends heavily on memory performance.
Etched says it redesigned hardware for both stages. For prefill, it built a low-voltage chip intended to pack in more transistors while avoiding typical heat problems in high-end AI chips. For decode, it created a new memory and interconnect approach called “cluster-scale memory,” designed to let many chips access a shared memory pool at very low latency.
Jane Street’s role matters
Jane Street said it tested the chip and was pleased with the early results. The firm also said Etched’s inference approach provides the precision needed for its most demanding workloads, and that it now has its own rack running in its data center.
That customer validation is important for a hardware startup. AI chips and systems face a long path from design to real deployment: manufacturing, cooling, networking, software compatibility, reliability, and customer operations all matter. A financial trading firm such as Jane Street is generally associated with demanding technical infrastructure, so its decision to buy, deploy, and lead the round gives Etched a stronger commercial proof point than investor enthusiasm alone.
Etched is also trying to move past an earlier perception of its business. The company originally intended to etch a particular model into its chips, implying a highly specialized design for one frontier model. It now says that is no longer the case and that its systems can run any frontier model. That change is central to the investment case, because buyers are less likely to commit to hardware that could become tied to one fast-changing model family.
What the deal says about the AI market
The financing reflects a broader shift in AI infrastructure spending. Training frontier models has consumed enormous GPU capacity, but once AI products are used at scale, inference becomes a recurring cost. Every search query, coding request, office task, trading workflow, or chatbot interaction can trigger inference. Lower latency and lower cost per generated token are therefore becoming strategic priorities.
Etched’s challenge is to prove that its combination of low-voltage prefill chips and cluster-scale memory can work beyond early tests and one high-profile customer rack. The company will have to show performance, precision, and cost advantages across multiple models and larger deployments, while competing against entrenched GPU ecosystems and mature software stacks.
For now, the round positions Etched as one of the most closely watched challengers in AI inference hardware. Its next test will be whether Jane Street’s deployment becomes an isolated milestone or the first example of a repeatable system business.