Da Mo warns that $1.6 trillion AI infrastructure investment faces multiple risks: the United States will still lack 32 gigawatts of electricity by 2028 According to Beating AI News, Morgan Stanley categorizes the core risks facing AI infrastructure construction in the United States as "people, electricity, and politics". It believes that local government tightening approvals, supply and demand reversal leading to a simultaneous decline in quantity and price, and power shortages may jointly impact AI infrastructure investment plans worth up to $1.6 trillion. At the political level, the selection of data center locations, grid connection applications, electricity pricing mechanisms, and tax incentives are mainly influenced by state and local governments. As the midterm elections in the United States approach, restrictions on data center construction may become a significant factor affecting the sentiment of the AI industry. The report points out that the cancellation of relevant tax incentives has become a cross party issue, and the regulatory decisions of the Public Utilities Commission may also have long-term impacts on project implementation. In terms of supply and demand, Da Mo warns that the current shortage of AI computing power is driving companies to purchase in advance and accept premiums. However, once supply catches up with demand or demand growth slows down, upstream products such as chips and memory may face a simultaneous decline in price and shipment volume, putting pressure on profit margins. Even if cloud vendors' capital expenditures only enter the platform phase, it may weaken the pricing power of suppliers. Electricity is currently the most direct constraint. The Great Wall Model shows that even with power supply options such as nuclear energy, gas turbines, and fuel cells included, the United States may still face a net power shortfall of 32 gigawatts by 2028, and the shortfall may widen to about 72 gigawatts by 2029. Delay in power delivery may result in cloud vendors purchasing GPUs that cannot be put into use in a timely manner, leading to equipment hoarding and inventory accumulation. Da Mo still maintains a constructive judgment on AI infrastructure, believing that power shortages are more likely to cause project delays rather than complete demand disappearance. But the bank warns that if political resistance, power bottlenecks, and supply-demand reversals overlap, it may lead to hindered chip procurement, a decline in upstream profit margins, and even reverse the entire AI capital expenditure cycle. [Original link]