Etched, an AI hardware startup, announced on Tuesday that it has raised an additional $700 million, pushing its valuation to $21 billion. This latest funding round was led by Jane Street, a prominent quantitative trading firm, which invested after testing and purchasing Etched's AI cluster systems.
The company's valuation has seen a rapid increase, rising from $10.3 billion in July to $21 billion in August. Etched delivers its AI technology as full systems, which it refers to as "frontier inference clusters."
According to Etched co-founder and COO Robert Wachen, the company has developed two new components to accelerate AI inference. These include a prefill chip designed for low-voltage operation to process more tokens faster, and a new type of memory and interconnect called cluster-scale memory for the decode process. This cluster-scale memory is intended to allow multiple chips to connect and use a shared memory pool at high speeds and low latency.
Jane Street stated in a blog post that it tested the chip and was pleased with the early results, noting that Etched's approach to inference provides the precision needed for demanding workloads. The firm now has its own Etched rack running in its data centre.