AI data centres amplify urban heat islands: systemic energy waste and infrastructure inequity drive localised climate disruption

Indigenous knowledge systems, such as the Māori concept of 'mauri' (life force) or the Andean practice of 'ayni' (reciprocal balance), frame heat as a relational phenomenon tied to ecological harmony rather than a mechanical byproduct. These perspectives highlight how industrial heat disrupts cultural landscapes and spiritual sites, yet they are systematically excluded from climate policy. Traditional cooling methods, like evaporative cooling in the Middle East or earth-sheltered homes in North America, offer low-energy alternatives that challenge the tech industry’s reliance on high-GWP refrigerants. The omission of these systems in favour of AI-driven solutions reflects a broader erasure of Indigenous climate adaptation strategies.

The heat generated by data centres echoes historical patterns of industrial pollution, from 19th-century textile mills to mid-20th-century steel plants, where localised warming was a visible externality of economic growth. Urban heat islands were first documented in 1818 by Luke Howard, who linked them to the built environment’s heat-absorbing properties—long before data centres existed. The post-WWII suburbanisation of the U.S. and Europe entrenched zoning laws that concentrated industrial heat in marginalised neighbourhoods, a legacy now replicated in the siting of data centres near low-income housing. The current AI boom mirrors the early internet era’s unchecked energy demands, when data centres were built without thermal regulation standards.

In Japan, 'mottainai' (waste not, want not) principles challenge the AI industry’s energy profligacy, advocating for reuse and efficiency over perpetual expansion. African cities like Nairobi and Lagos, where data centres are proliferating, face heat stress exacerbated by unreliable grids and informal settlements—highlighting how digital infrastructure deepens existing inequities. In the Arctic, data centre heat is melting permafrost, threatening Indigenous livelihoods, yet Arctic Indigenous groups like the Inuit have no seat at the table in global AI governance debates. These cross-cultural examples reveal a global pattern: AI’s heat footprint is not a technical glitch but a symptom of extractive development models.

Data centres can emit heat equivalent to a small city, with cooling systems often consuming 30-50% of their total energy, creating a positive feedback loop of inefficiency. Studies show that urban heat islands can increase local temperatures by 2-8°C, with data centres contributing up to 9.1°C in extreme cases due to their concentrated heat output. The scientific consensus links this heat to increased energy poverty, respiratory illnesses, and heat-related mortality, particularly in vulnerable populations. Yet industry-funded research often downplays these risks, framing heat as a 'manageable' issue through carbon offsets or AI optimisation, rather than a structural failure of energy governance.

Artists like Agnes Denes have long explored heat as a metaphor for human excess, as in her 1982 work 'Wheatfield—A Confrontation,' which juxtaposed industrial heat with ecological fragility. In Hindu cosmology, 'Agni' (fire) is both a purifier and a destroyer, a duality mirrored in the ambivalence of AI’s heat—capable of warming homes or scorching communities. Indigenous Australian songlines encode heat as a living force, challenging the Western reduction of thermal energy to a calculable variable. These perspectives invite a spiritual reckoning with AI’s heat, framing it not as a bug but as a symptom of humanity’s alienation from ecological balance.

Scenario modelling by the International Energy Agency projects that data centres could consume 20% of global electricity by 2030, with heat emissions tripling if current trends continue. Alternative futures include district cooling networks, where waste heat is repurposed for urban heating, or AI-driven load balancing to reduce peak energy use. However, these solutions require radical shifts in energy democracy, such as municipal ownership of grids or Indigenous-led land-use planning. The most plausible near-term scenario is a patchwork of localised heat islands, where wealthy enclaves enjoy cooled spaces while marginalised communities suffer intensified heat stress.

Low-income communities of colour in the U.S. and Europe are 3-4 times more likely to live near data centres, facing higher rates of asthma and heat-related illnesses due to cumulative exposure to industrial pollutants and heat. In India, Dalit and Adivasi communities near data hubs in Hyderabad and Bengaluru report water shortages and crop failures linked to thermal pollution, yet their testimonies are excluded from policy discussions. Women, who are disproportionately responsible for household cooling in many cultures, bear the brunt of heat-related labour burdens, yet their perspectives are absent from AI ethics debates. The absence of these voices in the narrative reflects a broader pattern of epistemic violence in tech governance.