The power constraint for AGI is not easily resolved
The United States needs significant new power generation and transmission infrastructure for AGI. Building it is hard
Leopold Aschenbrenner argues that trillion-dollar computing clusters are essential for AGI, and he identifies power supply as the primary constraint. He argues that the necessary power infrastructure can be built within his aggressive timeline of 2028-2030. However, a closer examination reveals significant challenges that cast doubt on the feasibility of this timeline.
He notes1:
But it’s totally possible to [build the required power generation and transmission infrastructure] in the United States: we have abundant natural gas.
Powering a 10GW cluster would take only a few percent of US natural gas production and could be done rapidly.
Even the 100GW cluster is surprisingly doable.
Right now the Marcellus/Utica shale (around Pennsylvania) alone is producing around 36 billion cubic feet a day of gas; that would be enough to generate just under 150GW continuously with generators (and combined cycle power plants could output 250 GW due to their higher efficiency).
It would take about ~1200 new wells for the 100GW cluster. Each rig can drill roughly 3 wells per month, so 40 rigs (the current rig count in the Marcellus) could build up the production base for 100GW in less than a year.The Marcellus had a rig count of ~80 as recently as 2019 so it would not be taxing to add 40 rigs to build up the production base.
More generally, US natural gas production has more than doubled in a decade; simply continuing that trend could power multiple trillion-dollar datacenters.
The harder part would be building enough generators/turbines; this wouldn’t be trivial, but it seems doable with about $100B of capex for 100GW of natural gas power plants. Combined cycle plants can be built in about two years; the timeline for generators would be even shorter still.
While his calculations may be accurate, several critical challenges undermine his optimistic outlook. Historical precedents, such as the decades-long completion of the US interstate highway system, demonstrate that even well-funded and prioritized infrastructure projects often face significant delays. Regulatory and environmental hurdles, skilled labor shortages, and supply chain issues further complicate the rapid expansion of natural gas production and power plant construction.
He continues with his explanation:
Well-intentioned but rigid climate commitments (not just by the government, butgreen datacenter commitments by Microsoft, Google, Amazon, and so on) stand in the way of the obvious, fast solution. At the very least, even if we won’t do natural gas, a broad deregulatory agenda would unlock the solar/batteries/SMR/geothermal megaprojects. Permitting, utility regulation, FERC regulation of transmission lines, and NEPA environmental review makes things that should take a few years take a decade or more. We don’t have that kind of time.
He says that we don’t have the time because he views trillion-dollar data centers as vital to the United States’ security:
We’re going to drive the AGI dataenters to the Middle East, under the thumb of brutal, capricious autocrats. I’d prefer clean energy too—but this is simply too important for US national security. We will need a new level of determination to make this happen. The power constraint can, must, and will be resolved.
It may be true that the power constraint can and must be resolved. It is far from obvious, however, that it will be resolved in a manner commensurate with his aggressive timeline for AGI.
Moreover, the economic feasibility and market dynamics of ramping up natural gas production amid price fluctuations cannot be overlooked. Additionally, the limitations of alternative energy sources like solar, batteries, SMRs, and geothermal present further challenges in achieving the required scalability within the proposed timeline.
While the urgency of developing AGI-related infrastructure is clear, the historical, logistical and regulatory challenges suggest that Aschenbrenner’s timeline is overly optimistic. A more realistic approach acknowledges these complexities and aims for gradual progress rather than rapid, large-scale implementation. The power constraint may indeed be resolved, but not within the narrow window Aschenbrenner envisions.
Note that in Aschenbrenner’s original essay, he uses indented bullets for some of the items on this list; Substack’s limited text formatting capabilities do not seem to allow me to create indented bullets. I don’t think that the formatting change alters the substance of his argument, however.