The stock price of VertiV (VRT.US), one of the liquid cooling technology solution providers for the GB200 AI GPU server, Nvidia (NVDA.US)'s most powerful AI GPU server, has skyrocketed by more than 600% since 2023, and the increase has reached 103% since 2024. According to Wall Street analysts, under the strong impetus of Nvidia, the absolute leader in the AI chip field, liquid cooling is expected to move from “optional” to “required” in the field of ultra-high performance AI servers, which means that the future market size of “liquid cooling” solutions will be extremely large, and in terms of stock price expectations, the upward path of liquid cooling industry leaders such as Vertiv may be far from over.
In terms of the latest results and performance expectations, Vertiv also handed over a performance that is very satisfying to the market, which suggests that demand for liquid cooling technology in global AI data centers has surged, while also showing from the side that global companies are still extremely strong in demand for Nvidia's AI GPUs. Recently, Nvidia's GB200 liquid cooling solution provider Vertiv's results showed that the company's total orders for the first quarter increased 60% year over year, and the backlog of orders at the end of the period reached 6.3 billion US dollars, a record high in one fell swoop. Net sales in Q1 were US$1,639 million, up 8% year on year, and adjusted operating profit reached US$249 million, up 42% year over year.
In addition to strong orders and sales for the first quarter, Vertiv also raised its 2024 full-year performance forecast at a pace that exceeded market expectations. The median sales value shows that it is expected to increase by about 12% year-on-year from strong 2023 sales, adjusted operating profit of 1,325 million to US$1,375 million, and is expected to increase by about 28% for the full year of 2023, when the median value is strong.
In China's A Shares, Invec (002837.SZ), a leader in liquid cooling technology, also handed over an extremely strong first-quarter results report. During the reporting period, Invec achieved revenue of 746 million yuan, an increase of 41.36% over the previous year. Net profit attributable to shareholders of listed companies was 61.9752 million yuan, an increase of 146.93% over the previous year. Net profit attributable to shareholders of listed companies after deducting non-recurring profit and loss was $54.3077 million, an increase of 169.65% over the previous year.
Looking ahead to the future prospects of liquid cooling, the penetration scale of liquid cooling solutions is expected to enter an “explosive growth” model from 2024. According to Dell'Oro Group's February 2024 estimated data, the agency expects the data center thermal management market size (air cooling+liquid cooling) to reach 12 billion US dollars in 2028. At that time, the liquid cooling scale is expected to reach 3.5 billion US dollars, accounting for nearly 1/3 of the total thermal management expenditure, accounting for less than 1/10 of the current expenditure.
IDC, an internationally renowned research institute, recently released a report stating that China's liquid-cooled server market will continue to grow rapidly in 2023. In 2023, China's liquid-cooled server market reached US$1.55 billion, an increase of 52.6% compared with 2022, of which more than 95% use cold plate liquid cooling solutions. IDC predicts that in 2023-2028, the compound annual growth rate of China's liquid-cooled server market will reach 45.8%, and the market size is expected to reach 10.2 billion US dollars in 2028.
Liquid cooling — gradually moving from an “optional” to a “required option” for AI server cooling modules
Currently, AI servers using Nvidia H100 AI GPUs around the world are diversifying in terms of cooling solution choices, but air cooling is still the mainstream choice. Although liquid cooling is gradually gaining popularity due to its advantages in high-performance computing, such as more effective thermal management and energy efficiency, the deployment of liquid cooled servers has not fully spread to all systems using Nvidia H100 GPUs.
In the era of Nvidia's new Blackwell architecture GPU (B100\ B200\ GB200 AI GPU), due to the surge in AI GPU performance, from a theoretical and technical point of view, the scale of air cooling has almost reached the limit of air cooling capacity, and the era of liquid cooling has begun. As liquid cooling goes from “optional” to “required” in the AI server field, it will greatly increase market space and become one of the important segments in the field of AI computing power. Overall, liquid cooling not only ensures efficient 24-hour uninterrupted operation of the AI GPU server at optimal performance, but also helps extend the service life of the hardware.
The Nvidia GB200 supercomputing server performance can be described as the existence of a “world's only” computing power system. The GB200 AI supercomputing system built by Nvidia is based on two B200 AI GPUs and the self-developed Grace CPU. The performance of inference workloads based on the Big Language Model (LLM) can instantly increase by 30 times. At the same time, compared with the previous Hopper architecture, GB200's cost and energy consumption are drastically reduced by about 25 times. On the GPT-3 LLM benchmark with 175 billion parameter levels, the GB200 has 7 times the inference performance of the H100 system and provides 4 times the training speed of the H100 system.
This powerful performance improvement means that the air-cooled cooling module is no longer sufficient to support the normal cooling operation of the computing power system. This is also an important factor for Nvidia to use liquid cooling solutions on a large scale in GB200 AI GPU servers mass-produced in September.
As AI and machine learning algorithms become more complex, the corresponding demand for AI computing power is also growing rapidly. In particular, when training large AI models or performing large-scale AI inference processes, AI servers require high-performance GPUs to handle these computation-intensive tasks. These high-performance AI GPUs (such as Nvidia's GB200) generate significant amounts of heat when running and require effective cooling solutions to maintain operational efficiency and hardware longevity. The liquid cooling system can transfer heat more quickly and effectively from heat sources such as GPUs to radiators, thereby reducing the possibility of heat accumulation, greatly reducing the possibility of transistor burnout, and keeping the GPU running at high performance for a long time.
In terms of technology routes, the mainstream opinion in the industry is that cold plate indirect liquid cooling is expected to be fully penetrated and promoted before direct liquid cooling. Liquid cooling systems can be divided into direct liquid cooling and indirect liquid cooling according to the contact method between liquid and hardware. Direct cooling includes immersion and spray types. Indirect liquid cooling is mainly a cold plate liquid cooling solution. The cold plate liquid cooling technology is mature. There is no need to change the shape of existing servers. The processing difficulty is low, the cost is low, and the cooling power consumption can meet the needs of AI servers, and it is expected to be promoted first.
According to a research report by well-known agency Markets And Markets, the global data center liquid cooling market is expected to grow from US$2.6 billion in 2023 to at least US$7.8 billion in 2028, with a compound annual growth rate of 24.4% during the forecast period. Markets And Markets said that due to the development of devices such as artificial intelligence servers, edge computing, and the Internet of Things (OT), compact and effective cooling solutions are needed. The advantage of liquid cooling is that it can effectively handle large amounts of data in challenging situations by cooling small devices and small servers. Overall, the data center liquid cooling market is mainly driven by hardware requirements to improve cooling efficiency, energy efficiency, scalability, sustainability, and higher performance GPUs under the strong demand of modern data centers to process heavy-volume data.
Wall Street analysts are generally optimistic that the huge scale of investment in artificial intelligence technology by global enterprises will support the continuous expansion of data center capacity. This can be described as a major benefit for Vertiv. Most of the company's revenue comes from sales of products such as data center power management and IT liquid cooling and hybrid cooling systems used in data centers. The company's main business focuses on providing power management and various cooling technologies for data centers around the world.
Vertiv is currently committed to developing advanced liquid cooling solutions for AI data centers. According to public information, the next-generation NVIDIA AI GPU accelerated data center advanced liquid cooling solution developed by Vertiv and AI chip leader Nvidia (NVDA.US) is expected to be suitable for the GB200. Vertiv's high energy density power and cooling solutions are designed to support Nvidia's next-generation GPUs to safely run the most compute-intensive AI workloads with optimal performance and high availability.
According to data compiled by the agency, Wall Street analysts gave Vertiv 8 “buy” ratings, 1 “hold” rating, and no “sell” rating. The consensus rating was “strong buy,” and the most optimistic target price reached 102 US dollars (closing at a record high of 97.940 US dollars on Thursday). Noah Kaye (Noah Kaye), an analyst from Oppenheimer & Co., emphasized that “AI megatrends” are expanding the potential market for AI data center capacity, and the Vertiv high-density computing market alone is expected to reach $25 billion by 2026.
This liquid cooling technology leader from China is favored by Wall Street giant Goldman Sachs
Wall Street bank Goldman Sachs believes that artificial intelligence, the “stock driving fuel” of the global stock market, is far from being exhausted. The agency said in the latest forecast report released recently that the global stock market is currently only in the first phase of the investment boom led by artificial intelligence. This boom will continue to expand to the second, third, and fourth stages, boosting more and more industries around the world.
“If Nvidia represents the first phase of the AI stock trading boom — the AI chip phase that benefits most directly, then the second phase will be for other companies around the world to help build infrastructure related to artificial intelligence.” the agency wrote. “The third phase is expected to be for companies incorporating artificial intelligence into their products to increase the scale of revenue, while the fourth phase is an overall increase in production efficiency related to artificial intelligence, and this expectation can be realized in many companies around the world.”
In the second phase of the AI investment boom, the focus is on companies involved in AI infrastructure construction other than Nvidia, including semiconductor equipment providers such as Asmack and applied materials, chip manufacturers, cloud service providers, data center REITs, data center hardware and equipment companies, software security stocks, and utility companies. At this stage, Goldman Sachs specifically mentioned a Chinese listed company in the research report, namely Shenzhen Envicool (Shenzhen Envicool), which focuses on precision liquid cooling technology for servers, data centers, and energy storage systems.
The well-known agency IDC predicts that in 2023-2028, the compound annual growth rate of China's liquid cooling server market will reach 45.8%, and the market size will reach 10.2 billion US dollars in 2028. According to IDC data, driven by industry needs and policies, the size of China's liquid cooling server market will further expand in 2023, and there will also be more and more partners participating in the liquid cooling ecosystem, indicating that the market's attitude towards liquid cooling solutions for data centers is very positive. As Chinese artificial intelligence companies and organizations demand more and more intelligent computing centers, whether in terms of construction or computing power supply, the energy consumption of IT equipment in such data centers has increased dramatically, and an efficient liquid cooling system is needed to maintain an appropriate operating temperature. Otherwise, it will pose a huge challenge to the cycle management and operation of large model products.
Liquid cooling technology is popular all over the world, suggesting that demand for Nvidia's AI GPUs is extremely strong
Liquid cooling leaders such as Vertiv and Invict have handed over extremely strong performance data, and analysts' bullish expectations for Vertiv are heating up, suggesting that global data centers, especially AI data centers, are experiencing a surge in demand for liquid cooling solutions. It also shows from the side that global companies are extremely strong in demand for Nvidia's AI GPUs based on the Hopper architecture and the newly released Blackwell architecture.
Goldman Sachs expects that the four major technology companies, Microsoft, Google, Amazon's AWS, and Facebook parent company Meta, will invest as much as 177 billion US dollars in cloud computing this year, far higher than the 119 billion US dollars last year, and will continue to increase to an astonishing 195 billion US dollars in 2025.
According to media reports, Microsoft and OpenAI are in detailed negotiations on planning a large-scale global data center project costing up to 100 billion US dollars. The project will include an AI supercomputer temporarily named “Stargate” (Stargate). This will be the largest supercomputing facility in a series of AI supercomputing infrastructures that the two leaders in the AI field plan to build within the next six years.
Needless to say, this Big Mac-level AI supercomputer will be equipped with the core hardware for “millions of computations” — Nvidia's continuously upgraded AI GPUs, which aim to provide powerful computing power for OpenAI's future more powerful GPT models and more disruptive AI applications than ChatGPT and Sora Wensheng Video.
Although the increase in demand for core hardware such as AI GPUs may stabilize as supply bottlenecks are gradually eliminated, the market for underlying hardware will continue to expand, and the shortage of Nvidia's high-performance AI GPUs may be difficult to completely ease in the next few years. This is also the important logic that Goldman Sachs and other Wall Street banks are optimistic that Nvidia will hit the $1,100 mark in the next year (Nvidia closed at $887.47 on Thursday).
In particular, the big AI model and the technical scenario AI software must face — that is, the software development side will inevitably continue to purchase or upgrade AI GPU systems under the impetus of the trend of updating and iteration, so the AI hardware market will still be extremely large in the next few years. According to the latest forecast from market research firm Gartner, the AI chip market size will increase 25.6% over the previous year to reach US$67.1 billion by 2024. It is estimated that by 2027, the AI chip market is expected to be more than double the size of 2023, reaching US$119.4 billion.