As AI data centres guzzle energy, can we harness their wasted heat?

Data centres used 485 TWh of electricity in 2024, or 1.5% of global supply, and their share is projected to reach 3% by 2030 as AI demand grows. At the University of Southern Denmark, Danfoss technology recovers 100% of the Bitten supercomputer’s waste heat for the municipal heating network. Sonderborg Hospital’s heat-recovery project is expected to pay for itself within two years and has made the hospital energy independent for nine months of the year.

Analysis

The investable implication is narrower than “AI heat becomes a new energy source.” Heat recovery can improve a data centre’s local economics and community acceptance, but it does not generate electricity or materially relax grid constraints; the heat is useful only when a nearby district-heating network or industrial user can absorb it, and temperature, seasonality, and heat-pump costs matter. This favors sites planned around heat offtake, not necessarily the lowest-cost sites already chosen for land or power.

Potential beneficiaries are thermal-management and heat-pump suppliers such as Danfoss, and, conditionally, infrastructure providers including Schneider Electric and Vertiv. The article gives no evidence of material orders, margins, or a scalable US retrofit market, so this is not yet an earnings catalyst for those companies. For META and MSFT, treat heat reuse as a possible permitting and social-license benefit at selected sites—not a meaningful offset to power demand or a basis for changing consolidated forecasts. The examples cited do not establish that either company has achieved a comparable financial benefit.

Near term, sentiment may favor efficiency narratives; over 1–3 months, watch project awards, utility interconnection queues, and local permitting. Over 6–18 months, the structural opportunity depends on district-heating buildout and standardized liquid-cooling adoption. Contrarian point: recovered heat may be counted as an emissions or energy benefit while leaving the binding constraint—reliable electricity supply—untouched. Thesis weakens if projects cannot secure contracted heat buyers, or if disclosed recovery economics fail to clear conventional cooling alternatives.

Key Decisions for Investors

• No directional META or MSFT trade on this article alone; the operational examples do not demonstrate company-specific, material earnings impact.
• Add Danfoss and thermal-infrastructure providers such as Schneider Electric and Vertiv to a watchlist, not a buy list. Seek disclosed project volumes, economics, and repeat orders before underwriting revenue or margin upside.
• For a 1–3 month catalyst check, monitor data-centre permitting and utility interconnection outcomes alongside heat-offtake agreements. Treat heat recovery as a local acceptance and efficiency offset, not a substitute for power capacity.
• Falsify the opportunity case if deployments lack nearby, year-round heat demand or if reported project economics do not support adoption; upgrade conviction only with evidence of repeatable contracts and measurable customer savings.