A new giant in the AI chip market: Startup Etched valued at $10.3 billion

Etched, a startup developing specialized chips for AI technologies, has caused a major stir among investors. Founded by three former Harvard students, the company raised $300 million in a Series C funding round, bringing its market valuation to $10.3 billion. This figure means the startup's value has doubled in just seven months. This is reported by Techcrunch.com reports .

The investment round was led by the renowned Sequoia Capital. Other major financial institutions participating in the process included Andreessen Horowitz, SK Hynix, Jane Street, and Diffusion Capital. According to TechCrunch, this deal was recorded as one of the highest-valued Series C investment rounds in the history of Sequoia Capital. The Etched project is also supported by prominent figures in the tech world, such as Peter Thiel, Andrej Karpathy, and Dylan Field.

The primary goal of Etched is to create chips specifically optimized for the "transformer" architecture that underpins modern AI systems like ChatGPT and Claude. Founders Gavin Uberti, Robert Wachen, and Chris Zhu are offering their products to customers not just as a chip, but as a complete system. Currently, the company has already secured $1 billion in orders, and the first systems are being tested by clients.

Technological superiority and innovative approach

Although many experts believe that Etched chips are designed only for a specific model, company representatives deny this. They emphasize that the systems can run any AI model without issues, including "Mixture of Experts" architectures like DeepSeek and Qwen, as well as new designs like Mamba. Even the fact that Google is working on similar Frozen v2 chips for its Gemini model confirms that the path chosen by Etched is correct.

Etched chips accelerate the AI inference process in two stages:

• Prefill: The process of understanding the request and context. Etched has significantly accelerated this stage using low-voltage technology.
• Decode: The response generation stage. For this, the company created a new type of memory called "Cluster Scale Memory," which ensures that multiple chips work together with extremely low latency.

According to company representatives, the low-voltage operating mode reduces chip overheating, allowing more transistors to be placed on a single die. As a result, users can achieve high speeds while significantly reducing computational costs.

Although the Etched startup has announced that it has successfully manufactured its chips at TSMC plants, competing with giants like NVIDIA in the market will not be easy. However, a capitalization exceeding $10 billion and a large package of orders indicate that this young team has become a serious player in the AI infrastructure market.