Jane Street Bets $700M on Etched's AI Chip After Testing

A month ago, the AI hardware startup Etched was worth $10.3 billion, which is a more resilient trajectory than software projects like the one seen after Riot canceled its fighting game. Today, its valuation has doubled to $21 billion, according to a report by TechCrunch, which also recently revealed its 2026 startup scout hit list, and whose flagship conference, TechCrunch Disrupt, has an approaching ticket discount deadline. That's a near-$11 billion leap in less than four weeks, a blistering pace even for the frothy AI sector. The rocket fuel is a new $700 million funding round led by an investor that doubles as its first major customer: Jane Street.

This isn't a venture capital firm taking a flyer, especially as massive bets on AI infrastructure are reshaping the market, as seen with the recent $3 billion Nscale IPO. It's a definitive, real-world validation from a quantitative trading giant that put Etched's chips to work. The round shows a new pattern for winning in AI infrastructure: sell a crucial, sophisticated customer first, then let them open their checkbook. As we reported in Ferrari Sells First Electric Car for Record $40 Million, a single elite client can redefine a market.

Jane Street Rewrites The Startup Funding Playbook

The typical path for a deep-tech startup involves a long, private R&D runway, followed by cautious pilots with early-access customers, and then a gradual commercial ramp. Etched just skipped straight to the finale.

Quantitative trading firms like Jane Street operate at the edge of computational performance, where microseconds translate directly to millions of dollars. They don't deploy unproven technology on a whim. According to Etched's announcement, Jane Street installed and tested the startup's first shipped "frontier inference cluster" system. The result? The firm was "pleased with the early results," and is now leading a massive funding round.

"We tested the chip and are pleased with the early results. Etched’s unique approach to inference delivers the precision we will need to support our most demanding workloads. We’re excited to now have our own rack running in our datacenter," the firm stated.

This sequence inverts the venture capital model. Jane Street isn't betting on slides or benchmarks. It's investing in a product it already uses and depends on for its "most demanding workloads." For other potential customers, this is as strong a signal as it gets.

What Convinced The World's Toughest Customer

Etched's core pitch is a complete, purpose-built system for AI inference, the process of running a trained model to generate answers, a process that has recently gained renewed scrutiny as OpenAI grapples with AI safety. Co-founder Robert Wachen explained the company designed two novel components to tackle the two main phases of inference.

The challenge: Modern AI inference has a split personality. The "prefill" phase is about understanding a prompt and is intensely compute-heavy. The "decode" phase is about generating the answer, token by token, and is overwhelmingly memory-intensive.

Etched's solution: The company built a low-voltage prefill chip that packs more transistors without overheating, enabling faster processing. For the decode phase, it developed a proprietary interconnect and memory architecture it calls cluster-scale memory. This allows multiple chips to act as a single, unified memory pool, promising higher speed and lower latency for generating long outputs.

The company, which now employs roughly 400 people drawn from giants like Nvidia and Google's TPU team, has also dispelled an early myth. Its systems are not hardwired for a single AI model. Wachen confirmed Etched's hardware can run "any frontier model," making it a flexible, general-purpose inference engine.

The Crushing Weight Of A $21 Billion Valuation

A $21 billion valuation is not a prize. It's a set of handcuffs made of expectations. The funding validates Etched's technology, but it also instantly raises the stakes to an almost unimaginable level.

• Scale or fail: With over $1 billion in signed orders already on its books beyond Jane Street, the immediate imperative is flawless execution. The company must transition from a successful prototype deployment to industrial-scale manufacturing and support.
• The competitor target: Etched is now the undisputed heavyweight in the pure-play inference chip arena. Its staggering valuation will force every other contender, from Groq to a host of stealth startups, to re-evaluate their own pitches and fundraising strategies. This dynamic is similar to the pressure felt in other high-stakes tech sectors, like when Groq Slashes Valuation By Half After Nvidia Deal.
• The next domino: Jane Street is the ultimate reference customer for low-latency, high-stakes environments like finance. The next critical step is proving this architecture works for the cloud giants and major enterprises that drive the bulk of AI spending. Who signs a contract next will determine if Etched is a niche precision tool or a mainstream challenger.

The race is no longer about proving the concept works. That's done. The race is now about proving the company can build a cathedral on the foundation of a single, perfect brick. For every other AI hardware startup, the playbook has changed: secure a marquee, technically elite customer who can become your lead investor. For Etched, the hard part starts now.

Why This Changes Everything

• A $10.7 billion valuation leap in under four weeks signals that real-world customer validation, not just venture capital hype, is becoming the primary driver of AI hardware valuations.
• Jane Street's dual role as lead investor and first major customer rewrites the startup funding playbook, proving that landing one elite, performance-critical client can be more valuable than years of gradual commercial ramp-up.
• The deal demonstrates that quantitative trading firms—where microseconds equal millions—are becoming the proving ground and financial engine for next-generation AI infrastructure startups.