Limitations In Electric Grid Interconnections Should Not Slow The Development Of AI Data Centers

ARK’s research suggests that power shortages are unlikely to impede the expansion of AI data centers and that the higher electricity costs associated with rapid development of AI data centers will not impact their profitability significantly. Recently, for example, Elon Musk used generators to power xAI's data center in Memphis, Tennessee, bypassing full grid interconnection.

While growth in electricity production globally has averaged ~2.7% at an annual rate for the past five years, ARK’s research estimates that the incremental demand from AI data centers will be 0.5 percentage points, pushing the growth in global electricity demand to 3.2% at a compound annual rate through 2030, as shown below.

Note: “kWh” denotes kilowatt-hour, a measure of energy usage. Source: ARK Investment Management LLC, 2024, based on data from The Energy Institute, as of July 26, 2024.¹ Forecasts are inherently limited and cannot be relied upon. For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

For perspective, electricity generation in China has increased 5.7% at an annual rate over the past five years. According to our estimates, in 2023 alone, China installed more electric generating capacity than will be necessary to meet the likely incremental global demand from AI data centers in 2030. Therefore, increasing the rate at which we build out the needed power should be achievable. Furthermore, the industry perked up when Constellation Energy announced² that it plans to revive the Three Mile Island nuclear plant—which is only ~837 MW—retired in 2019, for a new 20-year deal to support Microsoft’s data centers.

If we are correct, there will be many more announcements regarding the amount of new power being brought online.

Note: “GW” denotes gigawatt, a unit of power used in the field of electrical engineering and energy production, equal to one billion watts. Source: ARK Investment Management LLC, 2024, based on data from Climate Cooperation China 2023 and Reuters 2023,³ as of July 26, 2024. Forecasts are inherently limited and cannot be relied upon. For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

The time required to build new generation and distribution capacity should not be a limiting factor. According to ARK’s research, electricity accounts for ~9% of total AI data center costs, leaving ample room for companies to invest in expedited, non-grid power solutions without disrupting data center economics—especially given the high returns on investment that we expect from AI advancements. As Elon Musk recently noted, waiting for conventional suppliers and grid interconnection would have extended the development of xAI by at least 12 to 18 months. Bypassing the grid in part, xAI brought the world’s largest training cluster online in four months.

Source: Winton 2024, Musk 2024.⁴ For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

What type of investment in power generation is needed to meet the incremental electricity demand? ARK’s research suggests that, given the overnight capital cost of natural gas today, the incremental capital required would be ~$235 billion in 2030, roughly 6% of what ARK expects to be spent on AI hardware that year.

Natural gas plants typically take two years to build, making them a feasible and achievable solution within a short time span. Modular nuclear also could be a solution, as the steady power demand from datacenters matches well with nuclear plant performance characteristics. Although solar and batteries could be paired for uptime 24/7, the location and size of data centers make renewable unlikely as a sole solution, as shown below.

Note: "MW” denotes megawatt, a unit of power equal to one million watts. Source: ARK Investment Management LLC, 2024, based on data from Thunder Said Energy 2024, as of July 26, 2024.⁵ For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security.

While the requirement for data centers to be located in specific geographies could lead to pockets of power shortages, they could be avoided if companies were to build generation behind the meter while waiting longer for grid interconnection. Indeed, meaningful investment in non-grid-connected generation could lead to a more distributed energy footprint, and potentially provide the risk capital needed to reinvigorate the nuclear industry.