Business

AI-generated text

Etched raises $700M at $21B valuation to challenge Nvidia in inference chips

Etched announced a $700 million funding round led by Jane Street that values the custom-inference chip startup at $21 billion, up from $10.3 billion in July and $5 billion in December.

Etched raises $700M at $21B valuation to challenge Nvidia in inference chips

Etched announced it has raised an additional $700 million in a funding round that values the custom inference‑chip startup at $21 billion. The round was led by Jane Street after the quantitative firm tested and purchased the company’s hardware. Etched’s valuation was $5 billion in December and $10.3 billion in July, meaning it doubled in one month to reach the current level.

What Etched builds

Etched develops dedicated inference chips along with a prefill processor and a cluster‑scale memory system. The stack is designed to run frontier AI models — large language and other generative models at the leading edge — faster and more cheaply than general‑purpose GPUs.

Market context and Groq’s precedent

Two years ago Groq demonstrated that a dedicated inference architecture could outperform Nvidia on latency and throughput. Nvidia later signed a licensing deal, absorbed the technology, and brought Groq’s founders and core team in‑house, after which Groq pivoted toward an inference cloud offering.

Etched positions itself as a new independent player pursuing the same objective Groq began: delivering an alternative, purpose‑built hardware path for inference workloads that can compete with the dominant GPU suppliers.

Why this matters

The new funding and the $21 billion valuation signal that investors see significant upside in dedicated inference hardware as a route to better performance and lower running costs for generative AI workloads. The earlier outcome for Groq — becoming part of Nvidia — underscores the stakes: Etched’s backers are betting this iteration of the story will remain independent and reach a different conclusion.

Key points

  • Etched raised $700 million at a $21 billion valuation.
  • Jane Street led the round after testing and buying Etched’s hardware.
  • The company offers dedicated inference chips, a prefill processor, and a cluster‑scale memory system aimed at running frontier models more efficiently than GPUs.
  • Groq previously proved the concept but was ultimately absorbed by Nvidia; Etched is attempting to complete that same mission as an independent firm.

How Etched executes on product delivery and how the market responds will determine whether this funding round marks the start of a sustained competitive alternative to GPU dominance.