Bruce Schneier and co-author Nathan Sanders argue that local resistance to AI infrastructure misses the more consequential question of corporate concentration of power.
Summary
- Community opposition to AI data centres is legitimate but may suit AI companies, who can absorb project defeats while pursuing far larger economic prizes.
- AI companies are targeting entire industries — software development, legal services, healthcare and education — not merely data centre real estate.
- Well-capitalised data centre projects have demonstrated they can override local opposition through litigation and federal support.
- The ‘AI safety’ debate in US elections is partly a marketing exercise by rival AI firms, both of which benefit from regulatory narratives that entrench their dominance.
- Schneier and Sanders argue the central risk is corporate concentration of wealth and political influence, and that meaningful responses require taxation, state regulation and public AI alternatives.
The opposition that suits AI companies just fine
Bruce Schneier and Nathan Sanders, writing in The Guardian, make a pointed observation: AI companies may be quite comfortable with data centre opposition as the primary focus of public concern. These companies can absorb a meaningful fraction of defeated proposals and continue building where it counts. Meanwhile, organised civic energy that might otherwise scrutinise how AI products reshape entire industries — or challenge the political influence of trillion-dollar corporations — remains fixed on local planning disputes.
Local concerns are real, but limited in scope
The authors acknowledge that community objections to data centres are grounded in genuine grievances. Data centres consume significant land and energy, produce very few local jobs relative to other industrial facilities of comparable scale, and their carbon footprint could grow materially if AI usage accelerates. Lower-income communities bearing the heaviest burden of these trade-offs have reasonable grounds for frustration. The authors note, however, that energy costs and inflation are currently more visibly affected by other factors — including the US-Iran conflict — and that other emissions sources, such as building heating, dwarf AI’s current environmental footprint by a considerable margin.
Capital wins when it matters
The limits of local opposition are already visible. An OpenAI- and Oracle-backed facility in Saline township, Michigan, is proceeding with construction after the developer sued the town of roughly 3,000 residents and forced a settlement permitting the project. The Trump administration has signalled readiness to advance AI infrastructure by overriding state objections and using federal lands. Schneier and Sanders observe that data centre campaigns appear most effective against speculative, early-stage proposals — precisely those with the lowest likelihood of proceeding regardless.
The larger prize: whole industries
The authors argue that data centres represent a fraction of what AI companies actually seek. The infrastructure spend — approximately three-quarters of a trillion US dollars this year by American companies — is smaller than the existing enterprise software market alone. AI companies, in the authors’ framing, are pursuing the value created by entire sectors: enterprise software development, creative services, management consulting, legal services, medicine and education. Resistance focused on construction sites leaves those ambitions largely undisturbed.
When rivals fund both sides of a policy debate
A notable dynamic has emerged in US electoral politics. PACs linked to Anthropic and OpenAI — rival companies — spent millions in a recent New York congressional primary on opposing sides of the ‘AI safety’ debate. The authors suggest this is more marketing than principled policy advocacy. Both companies benefit from a public discourse centred on the mystique of powerful AI requiring careful stewardship, whether that means light federal oversight (favoured by OpenAI-aligned interests) or a heavier compliance framework (which suits Anthropic’s market positioning). Either framing entrenches incumbent players and crowds out structural debate about corporate power.
Infrastructure may also be a shorter-term problem than assumed
The authors offer one counterintuitive note for those primarily focused on data centres as a long-term threat: centralised computing demand may not sustain its current trajectory. Chinese AI labs have pursued technical approaches that make frontier-capable models smaller and cheaper. Open-weight models are increasingly being adapted to run on personal devices, and both Apple and Google support on-device AI inference. The authors suggest the current data centre buildout could resemble the fibre optic bubble of the early 2000s, with demand eventually shifting toward smaller, locally-run models.
A broader policy agenda
Schneier and Sanders propose that effective responses must go beyond opposing construction approvals. They advocate for state-level AI regulation, taxation of AI computation to allow the public to capture some economic value while internalising environmental costs, and support for a ‘Public AI’ movement — an alternative ecosystem developed under public control with a public benefit mandate. They also call on political organisers to reject AI companies’ framing of policy debates and to pursue structural limits on corporate influence, including public campaign financing.
Why it matters
For CISOs and security executives, the governance vacuum described here has direct operational relevance. AI systems are already embedded in security tooling, vendor supply chains and enterprise software — the very industries Schneier and Sanders identify as targets for AI capture. If regulatory frameworks remain underdeveloped because the policy conversation is consumed by infrastructure disputes and AI-company-funded electoral narratives, organisations will continue making procurement and deployment decisions in the absence of meaningful external guardrails. The concentration of AI capability in a small number of vendors also creates systemic dependency risk: as these companies deepen their foothold in critical business functions, the leverage they hold over enterprise customers grows. Understanding the political economy of AI is not a distraction for security leaders — it shapes the risk environment in which every AI-related decision is made.
What to do now
- Engage with state-level AI regulatory processes where they exist, and monitor emerging legislation — the authors identify state regulation as a meaningful lever that AI companies are actively working to preempt.
- Assess your organisation’s vendor concentration risk as AI companies expand into enterprise software, legal services and other critical functions; build this into third-party risk frameworks.
- When evaluating AI vendor commitments to ‘safety’ or ‘ethics’, examine the regulatory and commercial interests behind those positions rather than accepting them at face value.
- Monitor the Public AI movement and open-weight model developments as potential alternatives that reduce dependency on a small number of dominant commercial providers.
- Factor political and regulatory uncertainty — including the US federal government’s posture on AI infrastructure and state preemption — into your organisation’s AI strategy and risk planning.
