AI Chip Startup Etched Emerges With $5B Valuation and $1B in Booked Orders

Etched, a startup aiming to challenge Nvidia's dominance in AI hardware, announced significant progress this week: TSMC has completed manufacturing of its custom chip, and the company has already secured $1 billion in contracted orders for complete systems built around the technology.

The Product and Market Opportunity#

The startup's offering centers on what it calls "frontier inference clusters"—integrated bundles combining its chips with custom-engineered hardware racks and proprietary software. The package is designed to run AI model inference (the compute-heavy phase after a user enters a prompt) faster, cheaper, and with greater power efficiency than competing solutions, according to Etched's claims.

Inference has become a critical pain point for AI companies scaling to serve customers at production volumes. Unlike the initial model-training phase, inference workloads represent both the biggest computational bottleneck and highest operational cost for operators deploying large language models commercially—making any efficiency gains in this area extremely attractive to investors and operators alike.

Funding and Backing#

Etched has now raised $800 million in total capital, with its most recent $500 million round completed in December valuing the company at $5 billion. The investor roster includes prominent venture firms (VentureTech Alliance, Ribbit Capital) and quantitative trading shops (Jane Street, Hudson River Trading, Two Sigma), plus backing from several renowned AI researchers—Andrej Karpathy, Geoffrey Hinton, Fei-Fei Li, Arthur Mensch, and Scott Wu—as well as billionaires Stanley Druckenmiller and Peter Thiel.

From Rejection to Recognition#

The timing of Etched's public emergence contrasts sharply with its early struggles. CEO Gavin Uberti and president Robert Wachen, who both left Harvard to become Thiel fellows and co-found the company in 2022, have been discussing their chip ambitions since 2024. Yet in 2023, even with a detailed 30-page investment memo outlining why AI would require specialized chips beyond general-purpose GPUs, the pair faced nearly universal investor rejection. The company was operating month-to-month and running critically low on cash before sentiment shifted.

The funding landscape has since transformed dramatically. Inference-focused chip designers are now high-priority targets; Cerebras completed a major IPO earlier this year, while competitor Groq . Amazon, Google, and Microsoft are building their own proprietary AI chips in-house, and OpenAI unveiled its just last month.