AI backlash turns into a business risk for Big Tech

The AI trade just absorbed a variable that was not on the spreadsheet six months ago: political and consumer resistance that is starting to look like more than a polling footnote. Axios reported Saturday that what registered as culture-war noise is bleeding into territory equity analysts can no longer bracket away. Trust is softening. Local fights over data-centre sites are multiplying. For companies that pitch AI as a smooth upward curve, the next round of monetisation depends on permission, power prices and public tolerance — not only on whether the models beat a benchmark.

Markets have been slow to catch up to that shift. The bullish case rests on an assumption that compute scales faster than opposition organises, a bet that looked comfortable when the policy conversation was still abstract. Morgan Stanley analysts, cited by Axios, called public pushback a “binding constraint, particularly around data center buildout”. That locution pulls the conversation out of commentary and drops it into project-finance models. Slower permits, heavier utility concessions, a politician who spots an easy populist line — each can stretch capital schedules and make the buildout returns look patchier than consensus forecasts.

On the polling, there is not much daylight. A YouGov and The Economist poll found 71 per cent of Americans believe AI development is moving too fast. That covers 77 per cent of Democrats and 68 per cent of Republicans. The Axios story pulled out a sharper number: among people aged 14 to 29, just 18 per cent described themselves as hopeful about AI. Nobody should read that as an imminent demand shock, but it is a signal that the industry’s social licence is thinner than the share prices assume.

That the scepticism cuts across party lines matters for a concrete reason. A divided electorate can still produce a shared brake when communities conclude the downside lands locally while the upside accrues elsewhere. Resistance is more likely to surface in zoning fights, utility hearings, school-board policy and consumer-product adoption than in a federal bill — and to arrive county by county long before Washington acts. The standard investor playbook models regulation as a Capitol Hill event. This story suggests the friction shows up through planning commissions instead.

Embedding AI into ordinary workflows, consumer products and public infrastructure is still the commercial thesis. Avriel Epps, an assistant professor at the University of California, Riverside, told Axios that “what is not inevitable is that these technologies will be embedded in every aspect of our lives, become indispensable, or replace humans”. Markets do not need a full revolt for that observation to bite. Enough hesitation from users, workers and customers — enough to flatten adoption curves below what the bullish models forecast — is sufficient.

From polling to permits

Heatmap has tracked local resistance tied to data-centre cancellations, and Time reported on a broader populist backlash centred on AI’s electricity consumption alongside the suspicion that gains will concentrate among a narrow ownership class. Sentiment turns into cost at exactly that junction. Data centres are not abstract software. They pull power, water and land. All three inputs are regulated at the local level, which opens a politics structured around a simple question: why should a town carry the infrastructure burden so that tech groups can chase margins? That kind of fight can delay projects even when Washington stays permissive.

Markets have a habit of ignoring social backlash until it collides with physical infrastructure, and local data-centre fights warrant more attention than online complaints for a straightforward reason. They produce binary outcomes. A site is approved, delayed or cancelled. Once cancellations begin to cluster, analysts have to revisit timelines rather than merely adjust sentiment. Timelines, in capital-heavy themes, carry as much weight as the underlying technology.

Electricity provides the clearest bridge between public frustration and earnings risk. If households begin to see AI buildout as a source of grid strain or rising bills, utilities and state regulators will face pressure to slow approvals, demand more investment or redistribute costs. None of that kills the AI trade outright. All of it makes the timeline messier and more expensive. Infrastructure investors who have treated the asset class as a neutral conveyor belt may find it has become contested.

The political risk runs wider than siting fights. Data for Progress found voters increasingly expect AI to hurt the economy, a framing that hands regulators, attorneys-general and local officials a concise argument: slow the rollout, then sort out the labour market and ratepayer spillovers. A federal crackdown is not imminent. Boardrooms should still stop treating public scepticism as a PR nuisance and begin treating it as execution risk — particularly at companies whose AI thesis depends on rapid consumer adoption.

Labour politics sit inside that risk too. When public anxiety is channelled through jobs, wages or bargaining power, companies are more likely to face pressure to add human review, slow product changes or spend more on compliance and reassurance. Each of those costs is manageable in isolation. Taken together, they erode the tidy operating-leverage narrative that has supported much of the sector’s enthusiasm.

Why markets should care

The US contrast with the rest of the world is striking. Stanford HAI reported that 59 per cent of global respondents in 2025 said AI offered more benefits than drawbacks, up from 55 per cent in 2024. The operative word is global. Broad international sentiment can be warming on average even as a US backlash grows strong enough to disrupt valuations in American tech, utilities and infrastructure names. Investors have spent the past year arguing about chips, model quality and revenue capture. The next argument may be whether the social environment can physically accommodate the buildout those revenue forecasts assume.

In valuation terms, the risk is a rerating. Premium multiples do not simply require better products — they require faster adoption, cheap enough power and a permitting environment that does not force companies to bargain over every expansion. Weaken one leg and the story shifts from unstoppable platform shift to capital-intensive execution. That is a more familiar equity narrative, and familiar narratives rarely sustain extraordinary multiples for long.

Some executives are already framing agency — not inevitability — as the live question. Axios quoted Arun Bahl saying “some version of AI is inevitable … but we have choice”. The remark is not anti-technology. It is a reminder that adoption is negotiated. Workplaces push back, schools resist, cities litigate, customers balk — and when any of that happens, the industry does not get to skip the politics and jump to scale. Companies such as Superhuman may still capture productivity gains, but markets routinely overpay when they mistake technical possibility for frictionless commercial rollout.

For investors, the read-through matters because of where the next disappointment is most likely to form. AI optimism has sustained premium multiples on the assumption that demand, infrastructure and public acceptance will all rise together. The evidence Axios assembled suggests those three lines may diverge. Public trust can weaken while buildout costs climb. The sector can still grow in that world — just at a slower and more politically expensive pace than the cleanest bull case needs.

The story matters, in other words, beyond the headline. Voters will not suddenly cast ballots against algorithms. The real risk is that resistance begins to appear in the line items markets must eventually price: permit timetables, utility bills, compliance costs and the willingness of customers to let AI deeper into everyday decisions. Once the question moves from whether the models can improve to whether the public will allow this to scale, AI stops being only a technology trade. It becomes a classic business-risk story.