Etched Exits Stealth: $800M Raised, $1B in Contracts for Transformer Inference Chip

AI chip startup Etched emerged from stealth on June 30 with a set of numbers that demand attention: $800 million raised across multiple rounds, over $1 billion in signed customer contracts, a working chip manufactured on TSMC’s N4P process, and first production racks shipping this summer. The company’s Sohu chip is an ASIC designed exclusively for transformer inference, embedding the architecture directly into silicon rather than relying on general-purpose GPUs.

The latest $500 million round, led by Stripes and closed in December 2025 at a $5 billion valuation, includes strategic investment from VentureTech Alliance (a fund with ties to TSMC), Jane Street, Peter Thiel, Hudson River Trading, Jump Trading, Two Sigma, and Ribbit Capital. Angel investors include Geoffrey Hinton, Fei-Fei Li, Andrej Karpathy, and Stanley Druckenmiller. The company has 400+ employees drawn from Nvidia, Google TPU teams, Broadcom, SK Hynix, and TSMC.

Etched’s bet is architectural specificity: by hard-coding the transformer operation into silicon, the chip delivers what the company claims are state-of-the-art throughput, latency, and power efficiency for inference workloads. The company also built a full-stack rack system — circuit boards, cooling, networking, and software — rather than selling chips alone. Independent third-party benchmarks have not yet been published, making the next 6–12 months critical for validating the claims.

Why it matters: Etched represents the most credible challenge yet to Nvidia’s dominance in AI inference, with a working chip, foundry partnership, and billions in orders — but the proof will come when production racks are independently benchmarked later this year.