Etched's AI chip valuation quadruples in eight months as Jane Street bets big on inference speed

Etched, a startup building specialized AI inference hardware, has secured $700 million in new funding that values the company at $21 billion—a remarkable jump from its $10.3 billion valuation in July. The funding round, led by quantitative trading firm Jane Street, reflects extraordinary investor confidence in the startup's approach to accelerating how AI models generate answers after receiving a user prompt.

The valuation sprint#

The company's growth trajectory rivals the most explosive startup rallies in AI history. Etched stood at $5 billion just eight months ago in December, then climbed to $10.3 billion over the summer. The latest round nearly doubles that figure in four weeks—a pace that underscores how intensely the market is competing for AI infrastructure advantages.

Jane Street's decision to lead the round carries particular weight: the firm didn't simply invest; it tested Etched's hardware in its own infrastructure, validated the results, and has now deployed a full rack of Etched systems in its datacenter.

A two-chip strategy for inference#

Unlike traditional approaches that run all AI tasks on uniform hardware, Etched has designed separate components optimized for each phase of how AI models work. When a user submits a prompt, the system must first understand the input (the "prefill" phase), then generate the answer token-by-token (the "decode" phase).

Etched's prefill chip operates at lower voltage, allowing engineers to pack in more transistors without triggering the heat and power issues that plague conventional high-performance chips. This density enables faster token processing during the initial phase.

For the decode stage, Etched developed what it calls "cluster-scale memory"—a shared memory pool that allows many chips to communicate at extremely high speed and low latency. Together, these innovations target what COO Robert Wachen described to TechCrunch as the core promise: "higher speeds and lower costs."

Dispelling early doubts#

Etched has spent the past year fighting a lingering misconception from its earliest days—that the company customized chips to run only specific AI models. This misunderstanding has faded as the company has demonstrated that its systems support any frontier model. The arrival of heavyweight backers including Kleiner Perkins, Sequoia Capital, Andreessen Horowitz, Peter Thiel, Tiger Global, and others suggests confidence that the misconception no longer shapes investor thinking.