From the ultra-low-cost DeepSeek benchmarking ChatGPT's release, to Microsoft's global reduction of AI data center projects, and Alibaba Chairman Joseph Tsai warning of a bubble in U.S. AI infrastructure investment, multiple warning signals are flashing: the prosperity of AI data centers may soon face a cooling down. Data shows that since the beginning of the year, following the DeepSeek incident in January, Goldman Sachs' "Energy + AI" thematic investment portfolio has been in a downward trend. Financial blog ZeroHedge previously pointed out that with the emergence of more efficient large language models (LLMs), the market is forming a new trend of "doing more with less." Goldman Sachs analysts James Schneider, Michael Smith, and others released a report on Thursday, bringing forward their previous forecast for the peak time of global data center capacity from the end of 2026. In a report to clients, Schneider stated that he updated the global data center supply-demand model, mainly considering the impact of DeepSeek and the new capacity brought by projects like OpenAI's Stargate. In the new forecast, he has brought forward the peak utilization time of global data centers to 2025 (previously the end of 2026), while predicting that the supply-demand tension will gradually ease from now until 2027. Nevertheless, the utilization rate of data centers will remain above historical average levels. Schneider pointed out that there are three major uncertainties for future AI data centers: first, the monetization ability of consumer-facing AI services is weak; second, large AI infrastructure projects may lead to oversupply; third, the efficiency improvements brought by "small" LLMs aimed at enterprises. From the demand side, according to the report, Goldman Sachs' global technology team recently revised down the expected shipment volume of AI training servers, updating the demand growth for various data centers. This adjustment is related to the slowdown in AI training demand and the speed of proliferation of AI inference and corresponding data center workloads. As a result, Goldman Sachs has lowered the demand growth expectations most directly affected by AI for 2025 and 2026. Additionally, Goldman Sachs has made minor adjustments to historical data to better reflect actual changes in data center supply. These changes have led to an upward adjustment of the historical demand baseline, but the new demand for the next 18 months has been revised down, while the demand trend for 2027 and beyond remains largely unchanged. On the supply side, Goldman Sachs' model has also been adjusted accordingly, incorporating actual supply that has already been put into operation by the end of 2024, while adding previously untracked small data center operators. This has resulted in an upward adjustment of about 2GW in historical supply, including both corrections to the historical baseline and actual capacity growth. In the long term, Goldman Sachs expects that new data center supply coming online in 2030 will increase by 8%, mainly due to some validated new construction projects Despite adjustments to forecasts, analysts remain constructive on data center operators such as Digital Realty (DLR) and Equinix (EQIX), noting that after expectations for AI demand become more rational, the risk/reward profile for these companies has become more balanced. Regarding future risks, Goldman Sachs also conducted a quick survey of its clients, asking what the biggest challenge facing the AI theme in 2025 would be. The results showed that a quarter of respondents chose: "Efficiency Gains."