Demand is not the issue, computing power is! Morgan Stanley warns: AI supply bottlenecks may persist for years

As corporate AI applications continue to deepen, more and more companies are beginning to achieve quantifiable returns from their AI investments. However, Morgan Stanley (hereafter referred to as MS) believes that the shortage of computing power is becoming a key bottleneck limiting the further expansion of the AI industry, and constraints from power, labor, and political factors may keep this problem persisting for several years.

MS's American thematic research strategist Michelle Weaver said on Wednesday (December 12th), "Our supply is extremely insufficient. We see computing power becoming a constrained resource. Power bottlenecks, political bottlenecks, labor bottlenecks - these factors will constrain supply in the coming years."

She also pointed out that 25% of S&P 500 index component companies can now quantify the returns from AI investment, up from 14% a year ago, and data center financing is also very abundant, with NVIDIA and Wall Street financial institutions collaborating to raise $5 billion for AI infrastructure as an example.

Currently, AI data centers (AIDC) have seen a significant increase in demand for high-reliability backup power sources, with clear trends toward product upgrades such as high power, high voltage, and smart parallelization. Data centers have become the most important incremental direction for the diesel engine and generator set industry.

CICC issued a research report pointing out that the overall performance of the power equipment industry this year is steadily upward, prompting investors to focus on structural opportunities.

CICC believes that the initial signs of the global power big cycle have emerged, and the development of new power systems is a long-term, continuous process, with the possibility of maintaining momentum until 2030. Major listed companies' orders, revenue, and profits are still on a solid upward trajectory, and after adjustments, current valuations have entered the historically undervalued range. Related stocks have long-term investment value.