Nvidia Faces a Rival? 'Big Short' Investor Burry Names Etched as a Serious Threat

'Big Short' investor Michael Burry has once again targeted Nvidia (NVDA-US), naming AI chip startup Etched as a potentially significant competitor to the AI chip leader. On Tuesday (18th), Burry retweeted a Wall Street Journal report on Etched, stating bluntly, 'This is a serious competition for Nvidia,' instantly thrusting the young AI chip startup into the market spotlight.

Burry's mention of Etched comes just as the company announced the close of a $700 million funding round, with its valuation nearly doubling from $10.3 billion to $21 billion in less than a month. The round was led by quantitative trading firm Jane Street, with participation from Kleiner Perkins, Sequoia Capital, Andreessen Horowitz, and Tiger Global. Reuters notes that investors are betting on the rapid growth of AI inference demand—the computational workload of using trained AI models to generate real responses.

Etched, headquartered in San Jose, California, now employs over 400 people and has secured over $1 billion in customer contracts with both public and private AI companies and cloud service providers. Jane Street is not only the lead investor in this round but also Etched’s first customer, having already begun deploying the company’s technology.

One of Etched’s most market-watched advantages is its chip development and validation speed. According to the Wall Street Journal, Etched claims it took just 44 days from receiving test chips manufactured by TSMC (TSM-US) to running AI inference workloads. This timeframe is far shorter than the industry standard of six months or more, indicating Etched’s attempt to shorten the time from chip design to actual deployment through a highly integrated engineering team.

This speed advantage is closely tied to Etched’s talent strategy. In 2024, the company hired Brian Loiler as Vice President of Platforms, who previously spent 22 years at Nvidia, primarily overseeing the HGX and DGX server systems. After joining Etched, Loiler helped recruit about 12 engineers from Nvidia, some of whom reportedly rejected counter-offers with higher salaries from Nvidia. According to the Wall Street Journal, approximately 15% of Etched’s 400 employees previously worked at Nvidia.

For Nvidia, the threat from Etched does not lie in directly copying its general-purpose GPU model, but in targeting the fast-growing AI inference niche. While Nvidia’s current GPUs handle multiple workloads such as AI training and inference, Etched aims to optimize computation for the moment AI models generate actual answers through specialized hardware, reducing cost, power consumption, and latency.

Etched’s core product, Sohu, exemplifies this strategy. The company initially positioned Sohu as a chip dedicated to running AI models based on the Transformer architecture, the core framework widely used in large language models like ChatGPT. Compared to general-purpose GPUs, specialized chips sacrifice some flexibility in exchange for superior speed and energy efficiency in specific workloads.

As AI services enter large-scale commercialization, demand for inference computing is rapidly increasing. After large language models are trained, every user query, text generation, or task execution requires data centers to continuously perform inference. This means that as AI usage grows, inference could gradually become a major component of AI infrastructure spending, creating market space for specialized inference chip startups. Reuters points out that investors are particularly focused on efficiency metrics such as tokens processed per dollar and per watt.

Etched only completed a $300 million Series C funding round in July, at which time its valuation was $10.3 billion, led by Sequoia Capital, with participation from Andreessen Horowitz, Jane Street, Diffusion, and SK hynix (SKHY-US). Less than a month later, the company closed another $700 million round, with its valuation jumping directly to $21 billion—highlighting the astonishing capital enthusiasm in the AI infrastructure market.

More notably, Etched is no longer relying solely on investor market expectations. The company has secured over $1 billion in customer orders and has begun delivering chips and systems to customers. This sets Etched apart from some AI chip startups still in the product development phase. However, Reuters also cautions that the semiconductor industry has seen many technically excellent chip companies fail to establish successful business models, so high-value orders and valuations alone cannot fully prove Etched’s ability to challenge Nvidia.

Etched’s rise also reflects a shift in the competitive focus of the AI chip market. The generative AI wave initially drove demand for model training, making high-end GPUs the most critical computing hardware. Now, as the number and frequency of AI model usage rapidly increase, the computational resources consumed during the inference phase are expanding, prompting the market to seek cheaper, more energy-efficient specialized solutions than general-purpose GPUs.

For Nvidia, its biggest advantage remains its complete CUDA software ecosystem, vast developer community, GPU product lines, and an end-to-end platform extending from chips to servers and data centers. Even if Etched demonstrates speed or energy efficiency advantages in specific inference workloads, large-scale replacement of Nvidia would require overcoming challenges in software compatibility, mass production, supply chain, customer migration costs, and long-term reliability.

Nevertheless, Burry’s comments carry symbolic market significance. Over recent months, he has consistently questioned the AI infrastructure boom and Nvidia’s market valuation, arguing that the sustainability of AI capital expenditure and large cloud providers’ demand for Nvidia chips may be overestimated. He has previously publicly criticized capital cycles within the AI industry and transaction arrangements among large enterprises, maintaining strong skepticism about Nvidia’s valuation and growth prospects amid the AI frenzy.

Burry now viewing Etched as a 'serious competitor' prompts the market to reconsider how solid Nvidia’s dominance in the AI chip space truly is. If Etched can convert its 44-day chip validation engineering efficiency into stable mass production and turn its $1 billion in orders into real revenue, specialized AI inference chips could become a significant alternative to general-purpose GPUs.

However, Etched’s rapidly rising valuation also means the market has already priced in very high growth expectations. From $10.3 billion in July to $21 billion in August, the company’s valuation doubled in less than a month. Going forward, it must prove this valuation is justified through actual chip performance, customer deployment, shipment volume, and revenue growth. Reuters cites Michael Ashley Schulman, partner at Cerity Partners, who notes that while Etched has impressive technology and strong customer interest, the semiconductor industry is full of cases where 'technically excellent companies failed to become successful businesses.'

Therefore, Etched is currently better viewed as a potential major competitor to Nvidia, rather than an opponent already shaking its leadership. The real key will be how large the AI inference market grows and whether the cost and energy efficiency advantages of specialized chips over Nvidia’s general-purpose GPUs are sufficient to convince major AI companies and cloud service providers to change their existing hardware architectures.

On the other hand, capital competition in the AI chip market is intensifying. Etched’s latest $700 million round, combined with the $300 million raised in July, means the company has raised at least $1 billion in a short period. Coupled with over $1 billion in customer contracts, Etched has transitioned from a tech-concept startup to a phase of simultaneous expansion in capital, talent, and customers.

For investors, Etched’s case highlights that the next phase of AI hardware market investment themes may shift from 'who can provide the strongest GPU' to 'who can execute the most AI inference at the lowest cost and lowest energy consumption.' If this industry trend continues, Nvidia may face not only competition from other GPU suppliers but also an increasing number of specialized chips designed for specific AI workloads.

As of Tuesday, when the related reports were published, Nvidia’s stock price fell 2.24%, closing at $219.97. While Burry’s naming of Etched is not enough in the short term to alter Nvidia’s dominant position in the AI chip market, with Etched’s soaring valuation, customer orders surpassing $1 billion, and a growing number of former Nvidia employees joining, this AI chip startup is gradually moving from the market’s periphery into a competitive landscape that Wall Street must now take seriously.