Goldman Sachs Research: AI Data Centers Moving Towards the Megawatt Age, Liquid Cooling Technology Key Competitive Race Track

Goldman Sachs Group, Inc. released a research report stating that currently, companies and industry participants are optimistic about the overall demand trends for data centers. With rapid technological development, data centers are undergoing a transformation from traditional cloud architecture to specialized artificial intelligence (AI) infrastructure, with increased computing power density and liquid cooling technology becoming core trends in this transformation. Industry Trends: High-density computing power and liquid cooling technology becoming mainstream Goldman Sachs Group, Inc. stated that during the 2025 Data Center World Congress, many industry-leading companies are generally optimistic about the demand prospects for data centers and mentioned that many projects are already scheduled for the coming years. The report specifically emphasizes that as computing power density continues to increase, the transformation process from traditional cloud to infrastructure designed specifically for AI is accelerating. For example, it is expected that rack power in data centers will increase from the current 120 kilowatts (Blackwell architecture) to 600 kilowatts (Rubin architecture), and possibly even up to 1 megawatt in the future. This high power density makes liquid cooling technology a standard feature for future data centers, especially in new AI data centers. Goldman Sachs Group, Inc.'s Global Research Institute pointed out that liquid cooling technology not only significantly improves performance but also reduces costs, especially liquid-liquid cooling systems are preferred by hyperscale data centers due to their efficient heat dissipation characteristics. Enterprise Dynamics and Technological Roadmaps Goldman Sachs Group, Inc. analyzed in detail the layouts of several companies in the field of liquid cooling technology: nVent Electric (NVT.US): Introducing various liquid cooling solutions, including gas-liquid hybrid and full liquid cooling systems, with a focus on providing highly customized rack internal cooling designs for AI data centers in the future. Carrier Global Corporation (CARR.US): Planning to launch a 1 megawatt Cooling Distribution Unit (CDU) by the end of this year, expecting data center-related revenue to double from last year's $500 million to $1 billion this year. Vertiv Holdings (VRT.US): Its 2.3 megawatt liquid cooling system is favored by hyperscale data center customers, and the company is closely aligned with NVIDIA Corporation's technological roadmap to support customers transitioning to liquid cooling technology. Schneider Electric: Providing on-site power generation solutions including fuel cells and gas turbines through collaboration with private equity firms to meet the energy needs of data centers. Its subsidiary Motivair is expanding into the Middle East and Asian markets as a leader in liquid cooling, with 96% of its supply chain located in North America, which allows it to effectively resist the impact of global supply chain disruptions. Challenges and Controversies However, there is disagreement in the industry regarding the sustainability of demand. While companies are confident in long-term orders, Goldman Sachs Group, Inc. points out that short-term growth may be affected by customer structure and macroeconomic factors, such as a slowdown in year-on-year orders in 2024 for companies like ABB; existing AI applications (such as ChatGPT) are still not sufficient to absorb future capacity, requiring the development of new application scenarios. Additionally, there is disagreement among some customers about the timing of their investment decisions - whether to immediately adopt the latest technology (such as Blackwell architecture) or wait for the next generation design.