Morgan Stanley: The application of AI in enterprises is accelerating, but the supply of computing power remains severely insufficient, and the power shortfall may continue for several years.

As the application of artificial intelligence (AI) in enterprises continues to deepen, an increasing number of companies have begun to see quantifiable returns on their AI investments. However, Morgan Stanley believes that insufficient computing power is becoming a critical bottleneck for further expansion of the AI industry, and constraints related to electricity, labor, and political factors may allow this issue to persist in the coming years. Michelle Weaver, an American thematic research strategist at Morgan Stanley, stated in an interview on Wednesday that the market is still experiencing a significant supply-demand imbalance in computing power. "Computing power is becoming a constrained resource," and bottlenecks in electricity, politics, and labor will continue to limit supply expansion in the next few years. Weaver noted that the pace of AI adoption by businesses is accelerating, and the economic benefits brought by AI have become increasingly evident. Among the companies in the S&P 500 index, about 25% are currently able to quantify the actual returns from their AI investments, significantly up from 14% a year ago. This indicates that corporate spending on AI is gradually transitioning from early experimentation and infrastructure development to a stage that generates measurable commercial value. Meanwhile, funding support for AI infrastructure development is not lacking. Weaver cited the partnership between NVIDIA Corporation (NVDA.US) and Wall Street Financial Institutions, Inc. as an example, with related collaborative plans seeking up to $500 billion in financing for AI infrastructure, demonstrating that capital continues to flow substantially into data centers and computing power construction. However, ample capital does not necessarily mean that AI infrastructure can expand rapidly in sync. Weaver believes that the two main factors currently limiting the supply of computing power are the shortage of labor required for data center construction and the lack of electricity to power these data centers. AI data centers consume a vast amount of electricity, and the construction of new power supplies, transmission networks, and related infrastructure takes a considerable amount of time. Even considering innovative power solutions such as repurposing bitcoin mining sites and fuel cells, Weaver estimates that there is still an electricity supply gap of about 10% to 20% for AI infrastructure. This means that even if companies have enough capital to purchase AI chips and build data centers, they may still struggle to realize actual computing power due to insufficient electricity or a lack of construction personnel. Weaver expects that these structural constraints will keep computing power scarce and valuable in the coming years. In addition to electricity and labor, political factors are also beginning to emerge as new obstacles to the expansion of AI data centers. Weaver pointed out that with the U.S. midterm elections approaching, opposition to data center construction is rising in some areas. Local residents are mainly concerned that large-scale data centers may drive up electricity costs, increase water demand, and affect water quality, air quality, and the local environment. Although data center operators can, to some extent, respond to consumer concerns about electricity costs and environmental issues by adjusting construction and energy plans, Weaver anticipates that as the midterm elections enter their final stages, political controversies surrounding data center construction may further intensify. Overall, Morgan Stanley believes that corporate AI applications are gradually transitioning from the investment stage to generating actual returns, but the main constraints currently faced by the AI industry are no longer just funding or chips; they extend further to include electricity, labor, and the approval processes for data center construction along with political resistance. Until these bottlenecks are addressed, the supply of computing power may continue to lag behind rapidly growing demand.