For tech investors, a message from the US midterms is clear – data centres must redefine their sustainability, writes Alejandro Litovsky
As commentators unpack the landslide midterm results in the US, tech investors are acutely aware of just how prominently the local opposition to data centre infrastructure has figured in political campaigns.
It moved from localised grassroots complaints into a campaign pillar that attracted $173m (€155m) in ad spend and, according to exit polls, was a critical deciding factor for 56% of registered voters. Even as the financing of data centres comes under increased scrutiny, the midterms have shown how local social opposition can quickly translate into a political agenda.
On the face of it, having a data centre in your backyard is a tough sell for the tech industry. The enormous computing power they demand is being financed to operate for decades but depends on shared resources like freshwater that are already strained in many regions. Even before the elections results rolled in it has been clear that social consent had started to fray. Data Center Watch found that local opposition blocked or delayed around $130bn of US projects in the first quarter of 2026 alone, roughly matching the whole of 2025.
The issue extends beyond the US. Around the world, and particularly in Europe, social consent is also determining whether a site can be permitted, powered, expanded and eventually sold. The bottom line is that the AI boom increasingly depends on the permission of the communities where its infrastructure is located. In turn, this requires companies to demonstrate that those social, resource and ecological systems can cope with, or even benefit from, hosting data centres over the long term. For investors, companies, communities and local politicians, finding a way forward requires thinking about sustainability not just in terms of resource efficiency but also in terms of relationships and context.
New investment frameworks are moving in that direction. The European Commission’s A-to-G sustainability rating for EU data centres explicitly mandates disclosures regarding local water stress and grid flexibility. Even if a facility achieves an A grade for localised efficiency, its absolute volumetric claim on a constrained local aquifer or an over-allocated power grid can still trigger severe community blowback. The ultimate risk for companies, investors and communities remains the absolute availability of resources, not just the relative efficiency of the asset.
For projects being built today, the question local communities will increasingly pose is not just how efficient a facility is, but what it adds. The strongest social licences belong to projects that add local infrastructure or resource capacity, rather than buying an offset elsewhere or funding a loosely related community project.
A different approach is possible, and indeed an AI economy will require it. The most credible models address relationships and context through what we could call ‘shared resilience’. This applies to the four most contentious areas: power, water, heat and land. Practical examples that these approaches are viable already exist.
Power: pay for the grid you use and share the benefit. Google says it will cover “100% of the cost of power we use and any new infrastructure needs driven by our growth”. Microsoft made a similar pledge in January. Google’s Energy Impact Program goes further, funding no-cost home repairs, weatherisation and efficiency upgrades, as well as solar and battery storage at schools and community centres. In North and South Carolina, Google estimates annual savings of $545–$729 per participating home. The benefit is local and tied directly to household affordability.
Water: reduce the draw on the same local system. In Quincy, Washington, Microsoft helped fund a water reuse utility that treats and recirculates data centre cooling water, reducing reliance on potable water. At Middenmeer in the Netherlands, Microsoft captures rainwater to offset cooling needs. Interventions hydrologically connected to the site are far harder to dispute than a global replenishment total.
Heat: turn waste into a utility. In Odense, Denmark, surplus heat from Meta’s data centre feeds the city’s district heating network: enough to warm 6,900 homes, according to Meta; which could rise to more than 12,000 after expansion. In Hamina, Finland, Google supplies recovered heat free of charge to the municipal network, which the local utility expects to cover 80% of its annual demand. An unavoidable waste stream can become a community energy input.
Land: bind benefits to development rights. In Lancaster, Pennsylvania, what the World Resources Institute believes is the first public community benefits agreement for a data centre requires $20mn for sustainability and local economic initiatives, 100% clean energy, water-use caps and noise limits, backed by financial penalties. WRI warns that such agreements can be performative when commitments are vague; the strongest are measurable, enforceable and tied to the project’s approval.
These examples illustrate the idea of ‘shared resilience’, in which the data centre infrastructure strengthens the social and ecological systems it depends on. The temptation to call these initiatives ‘net positive’ should be resisted. Any facility can fund aquifer recharge, community solar or district heating and still impose a net burden if its demand drives fossil generation, its water use exceeds replenishment in the basin, or its grid costs fall on other customers. The correct test is not whether a company funds a positive project nearby, but whether the facility, plus its intervention, leaves the local system better off on the resources that matter.
Benefits that are additional, local, durable and proportionate to a facility’s marginal burden should feed directly into investment assumptions: permitting risk, expansion options, cost of capital and exit value. Over decades, a host community that gains from a data centre underpins its value while one that disproportionately bears its costs creates a liability that may not yet be fully priced.
The midterms are a dramatic example of the necessity of putting data centres’ shared resilience at the heart of an AI economy. Stated clearly, shared resilience is something that local residents and policy-makers will be more able to get behind, and one where companies and investors will recognise the full pricing of the risks and long-term value of the assets.
To read the latest IPE Real Assets magazine click here.
