Excess Heat from AI Data Centres: A Valuable Asset for Climate Change Mitigation
With the increasing demand for artificial intelligence, data centres—essential frameworks fueling this AI movement—are poised to become significant energy consumers. A report by the Advanced Research Projects Agency in the U.S. presents a creative solution: harnessing the surplus heat generated by these centres to address global warming. This analysis highlights two prominent uses for both economic and ecological advantages: direct air capture (DAC) of carbon and evaporative water treatment.
With a substantial rise in energy requirements projected, AI data centres are expected to draw 150GW by 2030. Notably, nearly all consumed energy is transformed into excess heat. Carlos Díaz-Marín, who is spearheading the research alongside Zachary Berquist, succinctly states, “Data centres are akin to the globe’s largest toasters, yet we aren’t producing any toast.”
The research considered the practicality of repurposing waste heat for DAC, which extracts carbon dioxide from the atmosphere, and for evaporative water purification, both of which necessitate substantial thermal energy. Through detailed modeling, Díaz-Marín and Berquist established that leveraging waste heat in data centres could potentially mitigate the emissions and water usage linked to AI data processing. Berquist adds, “Even if these data centres operate on natural gas, pairing them with DAC can still render them carbon negative.”
Despite encouraging findings, challenges persist in realizing this mutually beneficial relationship. Present sorbents for carbon capture demand regeneration temperatures of about 80°C, while emerging materials might function at around 60°C—approaching the limits of what available waste heat can offer. Díaz-Marín emphasizes the urgent need for advancements in sorbent technology to lower the required operational temperature.
Sanna Syri, a climate change mitigation specialist at Aalto University, highlights the significance of exploring all possible methods for capitalizing on data centre waste heat. With DAC currently at a fledgling 0.01Mt CO2/year deployment level, investigating wider applications is essential, though it heavily depends on elements like geographic circumstances, policy incentives, and technological progress.
The possibilities are appealing, yet achieving these initiatives necessitates additional research. However, Díaz-Marín and Berquist remain optimistic. Berquist reflects, “The benefits we can derive from AI will be positive, yet its current environmental impact is detrimental. There’s a chance to make both aspects beneficial.” Díaz-Marín imagines a future where utilizing AI tools like ChatGPT also aids in alleviating global warming.
As society confronts the rising energy demands of AI, transforming waste heat emerges not just as a route to sustainability but as a dual-purpose innovation promoting environmental responsibility alongside technological progress.