AI chip positive news! Super Micro Computer stated that it has shipped over 100,000 new liquid-cooled GPU solutions in the current quarter, leading to a nearly 18% increase during trading hours

A company announcement brought good news, triggering a surge in the popular artificial intelligence (AI) concept stock Super Micro Computer (Supermicro).

On Monday, October 7th, in the early trading session of the U.S. Eastern Time, Super Micro Computer, which produces AI training model servers, opened high and continued to rise. By the end of the morning session, it surged more than 17.6% at one point, marking the largest intraday increase since March 4th. It then maintained an increase of over 10% and ultimately closed up approximately 15.8% at $47.74, reaching a new closing high since August 27th and the largest closing increase since May 15th.

Commentators generally believe that the sharp rise in Super Micro Computer's stock price is attributed to the shipment data disclosed by the company on Monday, highlighting strong customer demand for AI servers.

In the announcement, Super Micro announced the launch of a complete liquid cooling solution, including a powerful Cooling Distribution Unit (CDU), cold plates, Cooling Distribution Manifolds (CDM), cooling towers, and end-to-end management software. The company stated that this solution can reduce hardware costs in data centers, as well as the costs of cooling infrastructure that typically needs to run continuously. Since June of this year, over 2,000 liquid-cooled racks have been delivered, with over 100,000 GPUs equipped with this solution shipped in a quarter. The announcement stated:

"Super Micro Computer has recently deployed over 100,000 GPUs with liquid cooling solutions (DLC) for some of the largest AI factories and other Cloud Service Providers (CSPs) in history. As the power required for AI and High-Performance Computing (HPC) workloads on each server approaches 12kW, liquid cooling is a more efficient choice to maintain the working temperature required for each GPU and CPU. A single AI rack now generates over 100kW of heat, which needs to be effectively removed from the data center. Liquid cooling at data center scale can significantly reduce the power requirements for a given cluster. The highest power consumption can be reduced by 40%, allowing users to deploy more AI servers within a fixed power range, increasing computing power and reducing training time for Large Language Models (LLMs), which is crucial for these large CSPs and AI factories."

What does the shipment of over 100,000 GPUs mentioned above mean? Some commentators suggest that if the average price of these GPUs is comparable to the $30,000 per chip price of NVIDIA, it could imply an order size in the tens of billions of dollars.

Analyst Woo Jin Ho stated that assuming 1,500 out of the 2,000 liquid-cooled racks mentioned by Super Micro Computer were shipped in September at a price of around $3 million per unit, the sales revenue for September would be approximately $4.5 billion, laying a solid foundation for achieving the midpoint of the guided sales range of $6.5 billion for Super Micro's fiscal first quarter of 2025 ending in September 2024.

The announcement on Monday temporarily eased investors' concerns. Investors have been worried in recent months about Super Micro Computer's declining profit margins and questionable financial data. As of the market close on Friday, October 4th, the stock price of Super Micro Computer had fallen by over 51% in the past three months Wall Street News mentioned that on August 28th this year, Super Micro Computer announced a delay in submitting its 10-K filing to the U.S. Securities and Exchange Commission (SEC), which is the company's annual report for the 2024 fiscal year ending on June 30, 2024. The delay is needed to allow additional time to complete the assessment of the design and effectiveness of the company's internal controls over financial reporting as of June 30.

Analysis suggests that the timing of this delay is quite delicate, as just the day before, well-known short-selling firm Hindenburg Research had announced a short position on Super Micro Computer's stock, alleging accounting irregularities and involving "undisclosed related-party transactions, sanctions, and export control failures."

Nearly a month later, on September 26th, media reports revealed that following Hindenburg Research's short report, the U.S. Department of Justice is investigating Super Micro Computer, with the investigation in its early stages. A prosecutor from the U.S. Attorney's Office in San Francisco recently contacted some individuals seeking information from a former employee who accused Super Micro Computer and its CEO of accounting misconduct in April this year.

On September 26th, Super Micro's stock fell by approximately 12.2% in a single day.

Analyst Ho stated that the delay in submitting the 10-K report by Super Micro and the U.S. Department of Justice's investigation may overshadow any "favorable AI server demand fundamentals."

"The lack of financial transparency may cast doubt on the sales momentum for its fiscal year 2025, which requires revenue to grow by 87%, the midpoint of its $26 to $30 billion revenue target range. The company's (Super Micro) advantage in rack-level design engineering and shorter delivery times have helped it secure key AI deals, but with financial signals of danger emerging, concerns from customers and suppliers may intensify."

Ho believes that Hindenburg's short report highlights many lingering risks associated with Super Micro, which are already known to the public, summarized in Super Micro's 10-K filings or previously encountered, such as delisting risks. It is currently inconclusive whether Hindenburg's investigation into Super Micro's financial operations is inaccurate, but it is also challenging to directly link them to financial misconduct at Super Micro. However, Hindenburg's report serves as a reminder that after experiencing exponential growth for several years, there is a need to enhance transparency and strengthen financial controls