Across the world, the race to build the physical backbone of artificial intelligence is colliding with something less predictable than supply chains or chip capacity: public patience. A new global poll, covering 15 large economies, suggests that opposition to expanding AI-related infrastructure—particularly data centres—is not only real, but in the United States it is unusually intense. The US, according to the findings, has the lowest support for infrastructure expansion among the countries surveyed, a result that points to a widening gap between the scale of AI investment and the social licence required to sustain it.
For governments and companies that have treated data centres as a largely technical problem—permitting, power connections, cooling, land use—the poll reads like a warning label. It implies that the next bottleneck in AI growth may be neither hardware nor software, but community acceptance. And unlike earlier waves of technology adoption, this one is arriving with visible construction sites, rising electricity demand, and a set of environmental and economic questions that are increasingly hard to dismiss as “local noise.”
The poll’s headline finding is stark: Americans show the lowest support for expanding infrastructure tied to AI growth. Yet the deeper story is what that low support appears to represent. In many places, data centres are no longer perceived as neutral utilities. They are seen as energy-intensive facilities that can reshape local grids, drive up costs, and alter neighbourhoods—sometimes without delivering equally tangible benefits to residents. The backlash, in other words, is not simply anti-technology. It is often anti-impact.
Why the US stands out
The United States has long been associated with rapid deployment of new technologies, from broadband to cloud computing. But the poll suggests that when AI infrastructure becomes physically unavoidable—when it requires new substations, transmission lines, water for cooling, and large-scale construction—support can fall faster than expected. The US result may reflect a combination of factors that reinforce each other.
First, the scale and speed of build-outs in some regions have made the transition feel abrupt. Data centres are not incremental upgrades; they are large, discrete projects that can transform a landscape quickly. When multiple facilities are proposed in the same area, residents may experience the cumulative effect rather than the individual one. That cumulative pressure can shift sentiment from cautious acceptance to outright resistance.
Second, the US energy debate is unusually charged. Electricity prices, grid reliability, and the pace of renewable build-out are all politically salient. Data centres sit at the centre of these debates because they are both electricity consumers and potential partners in grid modernization. If residents believe the facilities will strain local systems or increase costs without corresponding improvements, opposition becomes more likely. Even where operators argue that they will use efficient designs or secure power through contracts, the perception of demand growth can still dominate public opinion.
Third, the US has a strong tradition of local control over land use and permitting. That means community concerns can translate into delays, legal challenges, and political pressure more readily than in countries where planning decisions are more centralized. The poll’s finding of low support may therefore be partly a reflection of how quickly public sentiment becomes actionable.
Finally, there is the question of trust. AI infrastructure is often built by large corporations whose incentives are not always aligned with local priorities. Residents may worry about who benefits, who pays, and who is accountable if promised outcomes—jobs, tax revenue, sustainability measures—do not materialize. When trust is thin, even modest negative impacts can loom larger.
A backlash shaped by everyday concerns
Although the poll focuses on support for infrastructure expansion, the reasons behind that support appear to cluster around familiar themes that have become more urgent as AI demand accelerates.
Construction disruption is one of the most immediate drivers. Data centres require heavy equipment, road traffic, noise, and long timelines. Even when projects are eventually completed, the process can be disruptive enough to sour sentiment—especially if residents feel they were not adequately consulted early on.
Environmental footprint is another recurring concern. Data centres consume energy, and while operators often emphasize efficiency improvements, the public conversation frequently centers on total consumption rather than efficiency gains. Cooling methods also matter. In some climates, water use becomes a flashpoint; in others, heat discharge and air quality concerns can arise. The poll’s implication that opposition is influenced by environmental worries fits a broader pattern: people may accept the idea of AI, but resist the physical consequences of scaling it.
Community impact and local economics also play a role. Residents may fear that data centres will raise property values beyond affordability, change neighbourhood character, or divert resources away from schools and healthcare. Others may worry that the jobs created will be limited in number or concentrated in specialized roles that do not benefit local workers. Even when data centres bring employment, the distribution of benefits can be uneven, which can fuel resentment.
Then there is the question of fairness. Many communities have experienced energy price volatility or infrastructure strain. When new large consumers arrive, residents may ask why their own needs come second. If the benefits of AI are perceived as flowing primarily to distant shareholders or national tech hubs, local opposition can intensify.
What makes this wave different from earlier tech rollouts is that the infrastructure is both visible and measurable. People can see construction. They can feel traffic. They can track energy usage. They can read about emissions and water consumption. That makes it harder for companies to rely on abstract promises of innovation.
Support varies widely across countries
The poll’s broader message is that sentiment is not uniform. Support levels differ across the 15 economies, suggesting that national context matters. Some countries may have more established frameworks for integrating large infrastructure projects with grid planning and environmental safeguards. Others may have stronger public communication channels or more credible commitments to sustainability.
In some places, data centres may be framed as part of a national strategy for competitiveness and digital sovereignty. Where governments actively coordinate energy policy, permitting, and environmental standards, public acceptance may be higher. Conversely, where infrastructure planning is fragmented or where energy transitions are contested, data centre expansion can become a proxy battle for wider anxieties.
This variation matters because it suggests that backlash is not inevitable. It is shaped. Policy design, transparency, and the credibility of mitigation measures can influence whether communities view data centres as partners in modernization or as burdens imposed from above.
The “real-world constraint” problem
For leaders planning AI expansion, the poll signals a shift in how risk should be understood. For years, the dominant constraints have been technical: compute availability, power procurement, cooling capacity, and construction timelines. Those remain real. But public sentiment is now emerging as a parallel constraint—one that can affect permitting, project timelines, and reputational risk.
In practice, low support can translate into delays even when projects are technically feasible. Permitting processes can slow. Legal challenges can multiply. Local officials may face pressure to impose stricter conditions. Operators may need to redesign projects, invest in additional mitigation, or negotiate community benefit agreements. These steps can add cost and time, and they can reduce the predictability that investors and planners rely on.
The poll’s implication that the US gap is sharper than in other large economies suggests that the US may face a particularly difficult transition. That does not mean AI infrastructure cannot be built. It means the path to building it may require more than engineering excellence. It may require a new approach to legitimacy.
A unique take: the legitimacy gap is becoming an infrastructure issue
There is a tendency to treat public backlash as a communications problem—something solved with better messaging. But the poll points to something more structural: legitimacy is becoming part of infrastructure itself.
Data centres are increasingly embedded in the physical systems people depend on—electricity, water, transportation, and land. When those systems are stressed, communities naturally ask who is responsible and who benefits. If the answers are unclear, opposition grows. If the answers are credible and shared, support can rise.
That means companies and governments may need to think differently about what “success” looks like. It is not enough to meet regulatory requirements. They may need to demonstrate that expansion is compatible with local priorities: grid reliability, affordability, environmental protection, and community resilience.
In the US, where local governance and political polarization can amplify conflict, legitimacy may require more granular engagement. That could include transparent reporting on energy sourcing, emissions accounting, and water use. It could also include commitments that are specific enough to be verified—such as investments in grid upgrades that benefit the broader region, not just the facility.
Another dimension is the distribution of benefits. If communities perceive that they bear the costs while others capture the gains, support will remain low. Community benefit agreements, workforce development programs, and local tax arrangements can help, but only if they are designed to address the concerns that actually drive opposition. Generic promises may not work. Residents want to know what changes for them, not just what changes for the industry.
The energy question sits at the centre
Energy is the thread connecting many of the poll’s implied drivers. Data centres are often described as efficient relative to older models, and operators frequently highlight advances in cooling and workload management. Yet public opinion may still focus on total demand growth. Even if efficiency improves, the overall scale of AI workloads can overwhelm gains.
This creates a tension between two narratives. One narrative says: AI infrastructure is essential and can be powered responsibly. The other says: AI infrastructure is arriving faster than the grid and environmental safeguards can keep up.
Bridging that tension requires more than claims of efficiency. It requires credible integration with energy planning. That includes aligning data centre expansion with grid capacity, ensuring that power procurement does not simply shift emissions elsewhere, and making sure that local reliability improves rather than deteriorates.
If the poll reflects a backlash driven by energy and environmental footprint concerns, then the solution is not only to build more data centres. It is to build the energy systems that make them sustainable in the eyes of the public.
What happens next
The poll’s findings should be read as a near-term signal rather than a distant trend. As AI demand continues to grow, more projects will be proposed. In places where support is low, the friction will likely increase. That could lead to a patchwork of outcomes: some regions