AI on the Ballot (Stratechery Article 11-4-2026)
78/ 100United States
Summary
Ben Thompson argues that the 2026 US midterms, in which Democrats swept Congress and Gina Hinojosa won the Texas governorship pledging a moratorium on data-centre construction in the state holding over 25% of planned US capacity by 2030, mark a shift in what constrains AI: from compute, to memory, to power, and now to politics. Unlike globalisation, where affected towns had no say, building data centres requires permission, so voters who felt voiceless before are now blocking projects, with nearly every relevant local ballot measure opposing data centres winning except in North Dakota. Winners are hyperscalers with permitted sites; losers are neoclouds, Nvidia short term, and startups.
Highlights
The final constraint is political, and that may be the most imposing constraint of all.
Building data centers requires permission, which is to say that people actually have a say.
It’s just as likely we are looking at scarcity for much longer than we expected, which will have ripple effects on everything.
Contrarian Viewpoints
- Elections swing back and forth, and Thompson himself notes permanency is not declared; demand for AI's ends has historically tended to win over opposition to its means.
- Efficiency gains, such as Nvidia's power-constrained token output, and repurposing of existing permitted sites could ease the constraint without new approvals.
Investment Implications
- Overweight hyperscalers with existing permitted sites they can repurpose over neoclouds dependent on new permits and financing.
- Underweight neocloud and data-centre developers exposed to permitting risk, and favour edge-inference beneficiaries such as Apple.