Etched: $300M Series C Doubles Valuation to $10.3B in 7 Months

Etched raised $300M at a $10.3B valuation led by Sequoia. That's the short answer. The longer answer is more interesting.

Sequoia once passed on Etched. On July 23, 2026, the firm led what it calls the largest Series C in its history to buy its way into the same company — a startup that has never shipped a single production chip, whose entire pitch is that Nvidia's do-everything GPU is over-engineered for the one job that actually matters now: running transformer models fast and cheap.

Etched $300 Million Series C: Round Terms and Lead Investors

Etched closed a $300 million Series C on July 23, 2026, at a $10.3 billion post-money valuation, led by Sequoia Capital with participation from Andreessen Horowitz, SK Hynix, Jane Street, and Diffusion. Sequoia says it is the largest Series C the firm has ever led — a notable detail given that Sequoia had reportedly passed on backing Etched in an earlier round.

A Valuation That Doubled in Seven Months

Etched's price history is the fastest part of this story. The company raised $500 million led by Stripes in December 2025 at roughly a $5 billion valuation. Seven months later, it's worth $10.3 billion — a straight doubling, with no independent benchmark data and no commercial shipment yet in the market to justify the markup beyond investor conviction and signed paper.

What Sohu Actually Is: One Chip, One Job

Etched builds Sohu, an application-specific integrated circuit hard-wired to run transformer models — the architecture underneath essentially every production large language model today, from GPT to Claude to Gemini. That is a deliberate rejection of the Nvidia model: Nvidia's GPUs are general-purpose, built to handle training, gaming, scientific simulation, and dozens of model architectures on the same silicon. Sohu drops all of that flexibility to specialize in exactly one workload.

The company unveiled working silicon and a rack-scale inference system on June 30, alongside two newer components: a dedicated prefill chip and a cluster-scale memory system. First hardware shipments to customers are expected this summer, with a stated ambition to reach gigawatt-scale production capacity by 2027 — co-founder and president Robert Wachen put it bluntly: "We have a lot of work to do to get to Gigawatt scale."

Figures from TechCrunch, MLQ News, and company statements as of July 23, 2026.

The $1 Billion in Contracts — and What It Doesn't Prove Yet

Etched says it has more than $1 billion in signed customer contracts, with clients already testing working systems ahead of summer shipments. That is a real number, but it is a company-reported one: there are no independent benchmarks yet, no public list of named customers, and no shipped revenue on the books. The most plausible buyers — the hyperscalers and frontier labs burning the most on inference compute, think AWS, Microsoft, Meta, xAI, and OpenAI — have not been confirmed by Etched itself.

That gap between signed-contract value and delivered, benchmarked performance is exactly what a $10.3 billion price is underwriting. It is the same gap every hardware upstart with no shipping product has to close, and it is worth tracking against the inference economics playing out more broadly on the AI Spending dashboard.

Why Inference, Specifically, Is the Battleground

Training a frontier model happens once, or every few months. Inference — actually running the model for every user query, every API call, every agentic tool call — happens constantly and scales directly with usage. As AI products move from demos to daily-use infrastructure, the compute bill shifts from a one-time training run to a permanent, usage-linked operating cost. That's the shift Etched is underwriting: it isn't trying to out-train Nvidia, it's betting the inference bill becomes big enough, and stable enough in its shape, to justify chips that do only that one job extremely well.

It's the same thesis explored on our inference vs. training chips breakdown — and it's a thesis that inference-focused AI companies like Baseten and Fireworks have already turned into real revenue by renting out GPU capacity smartly rather than building custom silicon at all.

The Bull and Bear Case

Bull case: transformers have been the dominant AI architecture for years with no credible successor in sight, inference spend is compounding faster than training spend across the industry, and Nvidia's own margins on GPUs leave enormous room for a specialized competitor to undercut on cost-per-token. Sequoia writing its largest-ever Series C check, after having passed once, is a strong signal that institutional conviction has caught up with the founders' original thesis.

Bear case: Etched has shipped nothing yet, has no independent benchmarks, and is betting everything on one architecture holding its dominance through a 2027 gigawatt-scale buildout. Nvidia's moat has never really been raw silicon performance — it's CUDA and the software ecosystem built around it over fifteen years, which a faster chip alone doesn't automatically dislodge. A $10.3 billion valuation on pre-shipment paper is a bet on execution risk resolving cleanly, and hardware companies rarely execute cleanly on their first at-scale production run.

The Bottom Line

Etched is the clearest bet yet that AI compute is bifurcating into two distinct markets — training, where Nvidia's flexibility still wins, and inference, where a narrower, cheaper, purpose-built chip might not need to be flexible at all. Sequoia paying up for a company it once turned down tells you the firm thinks that bifurcation is real and imminent. Whether $10.3 billion is the right price depends entirely on what happens this summer, when Etched's promises turn into shipped racks, real customers, and — for the first time — numbers nobody at the company gets to choose how to present.