Etched exits stealth with $800M and $1B in chip contracts
What's the deal? AI chip startup Etched has come out of stealth with $800m raised and more than $1B in customer contracts.
The company has built its first inference racks after a successful A0 tapeout. Early customer tests show state-of-the-art throughput, latency, and power efficiency.
Its Series B was led by Stripes, with participation from Ribbit Capital, Radical VenturesDealroom has a profile for this one. Try Dealroom →, Positive SumDealroom has a profile for this one. Try Dealroom →, Primary, and Argo. Backers include Jane StreetDealroom has a profile for this one. Try Dealroom →, HRTDealroom has a profile for this one. Try Dealroom →, Two SigmaDealroom has a profile for this one. Try Dealroom →, and JumpDealroom has a profile for this one. Try Dealroom →, with strategic investment from VentureTech AllianceDealroom has a profile for this one. Try Dealroom →.
Why now? Etched's first racks ship this summer, targeting frontier models such as many-trillion-parameter mixture-of-experts (MoE) systems, long context, and agentic workloads.
To get there, the company unveiled two breakthroughs. Low-Voltage Inference runs its chip's math blocks at under half the voltage of most AI chips, promising multiple times the FLOPs density.
Cluster-Scale Memory creates a shared low-latency memory pool across chips, using a proprietary interconnect and an HBM/SRAM hybrid design to lift both throughput and speed.
The startup has assembled 400+ engineers from NVIDIA, Broadcom, SK HynixDealroom has a profile for this one. Try Dealroom →, TSMC, and more. It also built a 2MW datacenter in its office and opened a Taiwan factory for round-the-clock work.
What could go wrong? Etched is betting on a single architecture in a market dominated by NVIDIA, whose general-purpose chips serve every workload.
Specialised silicon can win on efficiency but risks obsolescence if model designs shift. Shipping at scale from a young supply chain is also unproven.
The signal: The AI race is moving from training to inference, where running models cheaply and fast decides who profits.
Investors are now backing purpose-built challengers that attack power and memory bottlenecks head-on. Etched's $800m round shows how much capital is flowing to anyone who can chip away at NVIDIA's lead.
Read more: LinkedIn
