In the worst case, approximately 12 GW of that capacity, planned for the years 2028 to 2030, could be canceled outright. That 12 GW is the 8% figure.
The bigger exposure sits one layer up. About 60 GW, roughly 40% of the total pipeline, is currently waiting on regulatory approvals and could face delays as a result.
Susquehanna’s cancellation estimate represents 20% of that pending 60 GW.
Why 2027 looks safe and 2028 does not
The timing matters as much as the size. Susquehanna expects 2027 to be a peak year for new data center installations, and that year appears largely insulated.
The reason is mundane but important: most of the projects slated for 2027 have already secured their permits.
The risk window opens after that. Projects targeting 2028 to 2030 are more likely to still need approvals, which puts them in the path of whatever rules emerge from the post-midterm landscape.
Susquehanna’s analysis ties that risk to the 2026 midterm elections. The concern is that potential Democratic leadership could make approval processes more complicated for new projects.
Dan Ives has offered a similar read. He warned that between 5% and 10% of the approximately 800 to 1,000 planned US data center projects may face risks following Democratic congressional gains.
Texas is the state to watch
If the regulatory risk has a center of gravity, it is Texas. The state accounts for 27% of the planned electricity capacity tied to these projects.
According to the analysis, Democratic gains there have raised concerns about more stringent environmental policies.
The backlash was already here
Susquehanna’s warning lands on top of an existing problem. Local opposition to data center expansions has already led to more than $260 billion in canceled investments this year.
What the Susquehanna analysis adds is a political layer on top of the local one. Neighborhood objections were one hurdle; a tougher approval regime at higher levels of government would be another.
What this means for the AI buildout
For investors, the headline percentage is less important than where the risk is concentrated. An 8% worst case spread evenly across the pipeline would be manageable. An 8% worst case clustered in the later years and in key states like Texas is a different calculation.
The near-term picture looks relatively stable. If 2027 is the peak installation year and most of those projects are permitted, then companies supplying that buildout may not feel much immediate pain.
There are a few things worth tracking from here. First, how much of the roughly 60 GW in pending approvals actually clears, and how quickly. Second, whether policy changes in Texas materialize, given the state’s outsized share of planned capacity. Third, whether community opposition keeps adding to the more than $260 billion in canceled investments recorded this year.
It is worth keeping the framing in perspective. Susquehanna described a worst-case scenario, not a base case, and Ives framed his range as projects that may encounter risk.