The valuation doubled to $21 billion in a month! “Big Short” Bury backs AI chip upstart Etched: Nvidia (NVDA.US) welcomes rival

The Zhitong Finance App learned that the “big short” Michael Burry (Michael Burry) publicly stated on Tuesday that an AI chip startup called Etched poses a threat to NVDA.US (NVDA.US). According to reports, Etched recently completed financing of 700 million US dollars, and its valuation doubled to 21 billion US dollars in the past month.

Burry retweeted a story about Etched on social networking platform X and wrote, “This is tough competition for Nvidia.”

Burry highlighted two key pieces of information mentioned in the report: Etched claims that after the test chip was retrieved from TSMC, it only took 44 days to successfully run the AI inference workload, and this process usually takes six months or more; second, the company recruited Brian Loiler (Brian Loiler), a senior engineer who worked at Nvidia for 23 years in 2024, and Loyler then took about a dozen engineers from Nvidia.

As one of the most famous Nvidia bears on Wall Street, Bury has emphasized many times before that the market is in an “AI bubble,” believes that chips and related AI stocks are seriously overvalued, and accuses Nvidia of boosting revenue through circular finance traders.

Just last week, he also called Nvidia's latest plan to partner with Wall Street Bank to send $500 billion to chip buyers a “signal of despair.”

Nvidia's stock price has been falling for three consecutive trading days, with a cumulative decline of about 2.5%. On the retail forum Stocktwits, investors' sentiment indicators for Nvidia were “bearish.”

So how sacred is the Etched named by Bury?

The valuation doubled in a month. What is the origin of the $21 billion Etched?

According to data, Etched was founded in 2022 and is headquartered in San Jose, California, USA. It was co-founded by three young Harvard University dropouts — Gavin Uberti, Robert Wachen, and Chris Zhu. The company focuses on developing a dedicated ASIC chip (called Sohu) designed for the Transformer architecture, with the goal of challenging Nvidia on the AI inference circuit.

What's even more surprising is that this startup, which has just brought its first chip to market, has attracted the world's top capital.

On August 18, local time, Etched announced the completion of a new round of financing of 700 million US dollars, led by quantitative trading giant Jane Street, followed by institutions such as Capcom Huaying, Sequoia Capital, Anderson Horowitz, Tiger Global, Bain Capital Ventures, and Blackstone. The post-investment valuation reached 21 billion US dollars. Just a month ago, the company's valuation at the time of Series C financing was only $10.3 billion.

According to reports, Jane Street is not only the lead investor in this round of financing, but also Etched's first customer. Jane Street said, “We tested this chip and were very happy with the initial results. Etched's unique inference approach provides the precision we need to support our most demanding workloads.”

Can Etched shake Nvidia by abandoning “omnipotence” and betting on “specialty”?

The reason why Etched is favored by capital stems from the “extreme focus” of its technical route. Unlike Nvidia GPUs, which take into account the comprehensiveness of graphics rendering, general computing, and AI training, Etched bets on AI inference (Inference) — the computational process in which an AI model receives user requests and generates responses.

According to reports, Etched invested tens of millions of dollars to test equipment and built its own 2 MW data center in the office building, while strengthening supply chain control to accelerate product development. These measures helped the company overcome the inference load in just 44 days after getting the TSMC test chip, far exceeding the average speed of the industry.

At the architectural level, Etched's design allows multiple chips to communicate and operate collaboratively like a single processor. The company claims that compared to Nvidia's latest Blackwell chips, its architecture can significantly reduce inter-chip communication time — an advantage that is particularly critical in large-scale cluster deployments. Etched is challenging Nvidia's dominant position in the AI chip market with this specialization path.

Additionally, Etched continues to actively recruit talent from Nvidia. Currently, about 15% of the company's approximately 400 employees have worked for Nvidia. This talent siphoning effect is quietly eroding Nvidia's barriers in the field of AI chips and may pose a threat to its long-term competitiveness.