Valuation Doubles in Under a Month

Etched, an AI inference chip startup, announced on August 18, 2026 a new $700 million funding round, pushing its valuation to $21 billion. The jump stands out for its speed: on July 23 of the same year, the company had closed a $300 million Series C valued at $10.3 billion. In under four weeks, the company's market value doubled.

Jane Street Goes From Customer to Round Leader

The most notable detail of the announcement is not just the size of the check, but who signed it: quantitative trading firm Jane Street, known for its high performance algorithmic trading operation, led the round after testing Etched's hardware in practice. The firm bought a rack from the manufacturer and already keeps it running inside its own data center, becoming both the startup's first confirmed customer and its lead investor at once, an uncommon form of validation in the AI hardware sector.

Who Else Is in the Round

Beyond Jane Street, the investor list includes heavyweight names in venture capital and private equity: Kleiner Perkins, Sequoia Capital, Andreessen Horowitz, Tiger Global, Bain Capital Ventures, plus investor Peter Thiel. The previous round, $300 million in July, had already included participation from Nvidia, reinforcing the crossover interest from major hardware players and venture capital in Etched's thesis.

What Etched Builds: Chips Optimized for Inference

Unlike manufacturers betting on chips specialized for a single model architecture, Etched repositioned its product to support multiple AI architectures, focusing on accelerating inference, that is, the moment when an already trained model answers a query, rather than training itself. According to founder and CEO Gavin Uberti, the company's first rack took three years to go from scratch to delivery, but the expectation is that the next production cycle will be significantly faster.

A Capital Race Showing No Sign of Slowing

Etched's fundraising pace reflects a broader pattern across the AI infrastructure sector in 2026: ever larger rounds, valuations doubling within weeks, and growing competition for inference capacity as companies shift generative model workloads into production at scale. An institutional customer like Jane Street stepping in as lead investor signals that appetite for hardware alternatives to traditional GPU chips remains strong, even as Nvidia holds onto market leadership.

Why It Matters for Brazilian Agencies and SMBs

The direct impact of a round like this does not reach a Brazilian marketing or customer service automation agency's daily operations immediately, but it moves an important piece of the chain that determines the future cost of AI inference, that is, how much it costs to run an already trained model answering production API calls. More competition among inference chip makers tends, over the medium term, to push down the per token cost charged by the model providers running on that hardware, including the ones Brazilian agencies already use. It is worth tracking this kind of infrastructure backstage news as a leading indicator of where API prices tend to head over the coming quarters.