Hidden concerns behind the AI boom: By 2028, the US power shortfall may be equivalent to 44 nuclear power plants

Hidden concerns behind the AI boom: By 2028, the US power shortfall may be equivalent to 44 nuclear power plants

Morgan Stanley recently released its latest report stating that the construction of artificial intelligence (AI) infrastructure in the United States is driving domestic electricity demand into a new phase, and that electricity supply capacity may become the key limiting factor for AI industry expansion.

The bank's strategist Stephen Byrd pointed out in a research report titled “Powering AI: Bitcoin Conversion, Business Models, a US Power Shortage and the Big Picture” that by 2028, the total electricity demand of US data centers is expected to reach about 69 gigawatts (GW). Of these, about 10 GW will come from data centers under construction, another 15 GW can be accessed through the existing grid, but about 44 GW of power gap will still remain.

This latest figure is further revised upward from Morgan Stanley’s December forecast (36 GW gap). The report notes that, in terms of nuclear power generation, 44 GW is equivalent to about 44 nuclear power plants.

The report mentions that the US Department of Energy’s Loan Programs Office has recently stated that it is prepared to provide hundreds of billions of dollars in financing to nuclear power projects, in order to promote the construction of clean energy capacity and ease potential electricity supply pressures.

Morgan Stanley believes that electricity supply shortages may impact the rollout and pace of AI-related investments. Estimates indicate that every additional 1 GW of data center capacity costs around $5 to $6 billion to build, and inadequate power access may lengthen the construction cycle of AI infrastructure.

Morgan Stanley emphasizes that the US currently does not have any new nuclear reactors under construction; considering that a nuclear power plant typically takes more than ten years to build, unless the US increases supply capacity in the short term through natural gas, fuel cells, and retrofitting existing facilities, it may not be able to support the rapid expansion of AI infrastructure.

Time to Power Solutions

To address this issue, Morgan Stanley proposes multiple “Time to Power” solutions, alternative measures that do not depend on the traditional grid connection process and can provide power more quickly. Assuming all these solutions are implemented, the US’s power gap by 2028 can be reduced to about 20%, or 13 GW, which is still equivalent to 13 nuclear plants’ generation.

The report lists several potential solutions:

  • Natural gas turbine projects could add about 15–20 GW of electricity;
  • Fuel cell company Bloom Energy could contribute 5–8 GW (if annual output rises to 3 GW, potential supply could be further expanded);
  • Direct electricity deals between existing nuclear plants and data centers could provide about 5–15 GW (excluding indirect offsets using additional natural gas, already counted in the above 20 GW natural gas turbine figure);
  • Additionally, Morgan Stanley estimates existing Bitcoin mining facilities already have large (over 100 MW) sites with completed grid connection agreements, totaling about 20 GW of potential capacity, which could be converted into 10–15 GW of actual supply.

Among these options, Morgan Stanley believes converting Bitcoin mines into AI data centers has significant advantages in execution speed and risk control, and may gain greater market recognition in the future. The report also notes that Bloom Energy’s fuel cell system is another reliable “quick power supply” method, expected to drive a rapid increase in the company’s shipments.

In addition to fuel cell and Bitcoin mine conversions, Morgan Stanley anticipates a diversified “Time to Power” transaction model will emerge, involving participation from independent power producers, turbine manufacturers, and energy companies.

Bitcoin Mine Transformation into Data Centers Gains Attention

With the rapid growth of AI computing power demand, Morgan Stanley pays particular attention to the trend of Bitcoin mines being converted into high-performance computing (HPC) data centers. The report notes that there are mainly two business models in the industry: The first is the “New Neocloud” model, represented by IREN, where mining companies purchase GPUs or TPUs, build their own data centers, and then lease the computing facilities short-term to large-scale cloud providers or corporate clients. For example, IREN signed a five-year lease with Microsoft.

The second is the “REIT Endgame” model, where miners build the “powered shell” (the infrastructure excluding chips and servers) and then sign long-term leases with cloud service companies. For example, APLD signed a 15-year lease with an undisclosed cloud provider.

Morgan Stanley believes that both models have considerable value creation potential and demonstrate the pathway for traditional cryptocurrency infrastructure to transform into the AI computing sector.

The report also provides reference data for valuations of Bitcoin mine conversions to data centers, showing that for large mining sites with stable grid access and installed capacity over 100 MW, the enterprise value per watt (EV/W) multiple varies greatly. Morgan Stanley notes that the lower the valuation multiple, the more attractive the conversion opportunity.

 

Risk Warning and DisclaimerThe market has risks, and investment should be prudent. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial situation or needs of any particular user. Users should consider whether any opinions, views or conclusions in this article fit their specific circumstances. If you invest based on this, you do so at your own risk.