Etched closes $5B round as Nvidia challenger preps first rack shipments

Etched, a startup building specialized AI inference hardware, has raised $5 billion as it races to ship its first rack-scale systems this summer and challenge Nvidia's dominance in AI compute.

The company designs chips, racks, software, and manufacturing methods together to run frontier AI models with what it claims is best-in-class throughput, latency, cost, and power efficiency for both prefill and decode workloads. Its approach stands in contrast to general-purpose GPU architectures, targeting instead the specific demands of modern inference — including trillion-parameter mixture-of-experts models, long-context windows, and agentic workloads.

This latest financing follows $800 million the company previously raised across four unannounced rounds, including a strategic investment from VentureTech Alliance. Etched disclosed the prior funding total on its website alongside the new capital raise.

The startup says it will use the capital to scale production and fulfill more than $1 billion in existing customer contracts. Its first racks are scheduled to ship this summer. To support around-the-clock engineering, Etched has opened a factory in Taiwan and built a data center, test house, and new product introduction lab in its San Jose headquarters.

Etched's first silicon, built on TSMC's N4P process, returned earlier this year and is now undergoing validation with customers. The company is led by co-founders Gavin Uberti and Robert Wachen, both Harvard Thiel Fellows, alongside CTO Mark Ross, formerly CTO of Cypress, and a team of more than 400 engineers drawn from Nvidia, Google TPU, Broadcom, SK Hynix, and TSMC.

The hardware architecture centers on two key innovations the company unveiled publicly: Low Voltage Inference (LVI), which runs math blocks at under half the voltage of conventional AI chips to sustain 80% or higher peak FLOPs without thermal throttling; and Cluster Scale Memory (CSM), a hybrid HBM/SRAM design using a proprietary ultra-low-latency interconnect that the company says eliminates the memory bottlenecks that limit decode speeds on existing HBM-based systems.

Etched's customers and partners include leading AI companies, cloud providers, and hyperscalers, though it did not name them. The company's advisory and investor network includes figures such as Geoffrey Hinton, Peter Thiel, Andrej Karpathy, and OpenAI's Noam Brown and Tal Broda, among others.

The funding places Etched among the best-capitalized challengers to Nvidia in the AI chip market. The startup is betting that vertical integration — from transistor design to liquid cooling to customer rack deployment — will let it capture share in a market where inference demand is projected to grow rapidly as AI models proliferate in production use.