Inference Chip Startup Etched Raises $700M, Valuation Doubles to $21B

Inference chip startup Etched has raised an additional $700 million in a funding round led by Jane Street, just under a month after announcing a $300 million fundraise. The new round more than doubles the company's valuation to $21 billion, reflecting surging investor appetite for specialized AI inference hardware as demand for efficient, low-latency computing continues to outpace general-purpose chip supply.

Jane Street decided to lead the round after testing Etched's hardware offering, with the firm stating it "was pleased with the early results" and is "excited to have our own rack running in our data center." The round also drew participation from Kleiner Perkins, Sequoia, Andreessen Horowitz, Tiger Global, Bain Capital Ventures, and Blackstone, according to the company.

Etched's valuation trajectory has been steep. The July 2026 raise had already valued the startup at $10.3 billion, a figure that was double its valuation from December 2025 following the close of a $500 million funding round. Founded in 2022 by Harvard University dropouts Gavin Uberti and Chris Zhu, the Cupertino-based company announced in June 2026 that it had developed a working inference chip and signed more than $1 billion in customer contracts.

The startup's chip is built around two custom-designed components that target the two distinct phases of AI inference: a compute-intensive prompt processing stage, known as 'prefill,' and a memory bandwidth-intensive output generation stage, called 'decode.' Produced on TSMC's N4P process technology, Etched says its chip operates at a "much lower voltage than any other AI chip," which reduces heat generation and allows for a higher transistor density.

The hardware also leverages an architecture that creates a shared low-latency memory pool across the entire scale-up domain. The company claims its proprietary ultra-low-latency, high-bandwidth interconnect enables "dramatically faster" memory access across chips, a critical factor for scaling inference workloads across large clusters.

Etched currently employs around 400 people at its San Jose, California, office, where it operates a 2MW data center. The startup has also recently opened a new 80,000 square-foot (7,430 sqm), 10MW facility in Milpitas, to the north of San Jose, signaling its intent to scale both hardware deployment and internal infrastructure.

"Scaling intelligence requires a new kind of inference hardware," Etched wrote in a LinkedIn post announcing the raise. "Our mission has never been more urgent." The latest funding round underscores a broader industry shift toward specialized inference solutions, as hyperscalers and enterprises alike seek to optimize cost and energy efficiency for AI workloads.