Public opposition to the construction of data centres has risen sharply, with a growing number of voters blaming the facilities for increasing electricity prices, according to a new analysis by Wolfe Research.
The investment research firm said public resistance has climbed to 53 per cent from 28 per cent just nine months earlier, making policy intervention one of the most significant near-term risks to the artificial intelligence (AI) capital expenditure cycle.
According to the report, the shift in public sentiment is already translating into policy action, with New York Governor Kathy Hochul imposing a one-year moratorium on permits for large new data centres. Wolfe Research also identified the Trump administration’s evolving restrictions on frontier AI models as another major policy risk facing the sector.
Polling conducted by Heatmap and Embold Research, cited in the report, found that voters now regard data centres as the leading cause of higher electricity costs. The facilities ranked ahead of oil and gas companies, the conflict with Iran, the Trump administration, ageing electricity infrastructure and utility providers.
Despite the growing political pressure, Wolfe Research does not expect a widespread wave of state-level bans on data centre developments.
The analysts noted that at least 15 U.S. states have introduced legislation proposing data centre moratoriums. However, none has been enacted into law. A similar proposal in Maine was vetoed by the governor, while comparable bills in New Hampshire and South Dakota were rejected by lawmakers. Other proposals have largely stalled without further legislative progress.
New York’s restrictions were introduced through an executive order directing the state’s Department of Environmental Conservation not to issue permits for new data centres with power demands of 50 megawatts or more. Wolfe Research observed that executive action could allow other states to adopt similar measures more quickly than through the legislative process.
Even if additional states impose restrictions, the firm believes data centre investment is more likely to shift geographically than decline altogether, with states such as Texas continuing to actively attract large-scale AI infrastructure projects.
On the federal front, Wolfe Research said the Trump administration’s traditionally hands-off approach to AI regulation is becoming increasingly interventionist as frontier AI models become more capable.
The report highlighted the administration’s collaboration with Anthropic, which granted the federal government early access to its Mythos model to help assess potential cybersecurity risks. It also pointed to subsequent export controls imposed on the Mythos and Fable family of models, describing the measures as an effective ban implemented without a clearly defined regulatory framework.
Separately, OpenAI agreed to limit the initial deployment of its GPT-5.6 models to a small group of trusted government partners before making the models more broadly available.
The analysts also referenced an executive order signed by President Donald Trump requiring covered frontier AI models to be shared with the government and designated trusted partners at least 30 days before public release. The implementation framework for the order was expected to be finalised by 1 August.
Looking ahead, Wolfe Research expects the White House to increase efforts to discourage the adoption of Chinese open-weight AI models. The analysts cited concerns surrounding Moonshot AI’s Kimi K3 model and allegations that it benefited from distillation techniques using leading U.S. frontier models.
The firm believes that any tighter restrictions on Chinese AI models would likely strengthen demand for American closed-model providers, potentially extending the current AI investment cycle.
While acknowledging that regulatory intervention is becoming more prominent, Wolfe Research concluded that policy measures are unlikely to significantly disrupt AI capital spending in the immediate future. However, it warned that regulatory constraints are expected to become increasingly influential as AI systems continue to advance in capability.
