The AI chip startup Etched announced on Monday, August 18, a $700 million Series D funding round, doubling its valuation to $21 billion in less than a month. The round was led by Jane Street, a Wall Street quantitative trading firm that tested Etched's hardware before becoming the company's first customer to receive a computing rack, according to a report from The Wall Street Journal. The investment is backed by established investors including Kleiner Perkins and Sequoia, which participated in previous rounds.
Etched was founded in 2022 by Gavin Uberti and Chris Zhu, two Harvard students who dropped out of university to build chips and memory components designed specifically to accelerate inference on artificial intelligence models without requiring traditional GPUs. Unlike many other AI chip startups that aim to build general-purpose GPUs to compete directly with NVIDIA, Etched has developed purpose-built hardware around the transformer model, the neural network architecture that powers systems like ChatGPT, Claude, and Gemini.
The $21 billion valuation for Etched reflects an extraordinary financial escalation. In July, the company had closed a $300 million Series C round with a $10.3 billion valuation — already double the valuation from December, when it closed a $500 million round. In less than a month, the company effectively doubled its market value, jumping from $10.3 billion to $21 billion, and raised an additional $700 million in capital.
Jane Street, one of the world's leading quantitative trading firms, not only led the round but also became the first customer to receive a computing rack from Etched. The firm tested Etched's hardware and decided to invest and become a customer — a rare combination that signals confidence both in the product and the team. The announcement that Jane Street had received its first rack marks a significant milestone: Etched has moved beyond being merely a hardware promise to becoming an actual supplier of AI computing infrastructure.
Etched's central thesis is that AI inference — the process of having a trained model generate responses from new inputs — is the dominant bottleneck in the cost of running artificial intelligence models. With the popularization of ChatGPT, Gemini, and other tools, the amount of inferences performed daily has grown exponentially, making inference efficiency as important as training capability.
Etched's approach starts from the principle that, rather than building general-purpose GPUs that try to do everything, it makes sense to create specialized hardware for inference on transformers. This co-designed hardware, developed alongside racks, software, and manufacturing methods, promises to deliver best-in-class throughput, latency, cost, and power efficiency for both prefill and decode workloads.
With a $21 billion valuation and only a year since its last $10.3 billion round, Etched is making a concentrated bet on AI inference. The AI chip market remains dominated by NVIDIA, which already has a massive installed base and a robust software ecosystem. The question is whether Etched's specialization will be enough to capture a significant share of this expanding market — or whether the company will need to demonstrate performance and cost advantages that genuinely displace customers from existing GPUs.
Sources: Unite.AI, SiliconANGLE, Dealroom
✓ Independent sources cross-checked and verified before publishing
