Etched puts working silicon and $1B in contracts behind inference ASICs
Original: Etched Emerges From Stealth With Working Chip, $800M Raised, and Over $1B in Customer Contracts View original →
Inference-specific AI chips now have a stronger proof point than a pitch deck. Etched says it has working silicon, more than $1 billion in signed customer contracts and $800 million raised, putting real manufacturing and demand claims behind its bet that transformer-heavy AI workloads need a different kind of infrastructure.
In its GlobeNewswire release, Etched says it achieved first-pass A0 silicon success on TSMC’s N4P process and is validating its first rack-scale product with customers. The systems are already running models including DeepSeek, Qwen, Mamba and Llama, and are designed for both prefill and decode workloads.
The financing is large even by AI infrastructure standards. Etched says it has raised $800 million across multiple rounds, including a $500 million financing in December at a $5 billion post-money valuation. The investor list includes VentureTech Alliance, Peter Thiel, Jane Street, Hudson River Trading, Jump Trading, Two Sigma, Stripes, Ribbit Capital, Radical Ventures, Primary VC, Positive Sum and several AI figures including Andrej Karpathy, Geoffrey Hinton, Fei-Fei Li, Arthur Mensch and Scott Wu.
The important framing is that Etched is not selling a lone chip story. The company says it is building rack-scale systems co-designed across the stack for throughput and latency, and that it has set up a Taiwan factory plus a data center, test house and NPI prototyping lab at its San Jose headquarters. That emphasis matters because AI inference constraints usually show up at system level: power, memory, cooling, networking, compiler support and deployment operations.
There is still plenty to prove. Signed contracts are not the same as broad production deployments, and the market will need evidence on real workload performance, software maturity, yield and reliability. But a working TSMC-produced chip combined with $1 billion in customer contracts makes Etched one of the more concrete challengers in the inference hardware race. The next test is whether its systems can move from validation to shipped capacity fast enough to matter before GPU incumbents and cloud providers compress the gap.
Related Articles
Enterprise AI bottlenecks are shifting from model access to operational control. NVIDIA says its internal Enterprise Inference Hub serves more than 100 model endpoints and processes trillions of tokens every week.
Meta says custom silicon is critical to scaling next-generation AI and has published a roadmap update for its MTIA family. The company says it accelerated development enough to release four generations in two years as model architectures keep changing faster than traditional chip cycles.
Together AI raised $800 million at an $8.3 billion valuation, a large bet that open-model infrastructure can undercut closed-model economics. The company says annual bookings topped $1.15 billion last quarter and plans to expand capacity about 50-fold over five years.