Circular solutions could ease infrastructure pressures amid AI ‘gold rush’ 

Academic experts and industry leaders have called for a shift towards circular infrastructure models as the demand for artificial intelligence places strain on global data centre supply chains. 

Speaking at Data Centre World 2026, Professor Fiona Charnley from the University of Exeter outlined how circular economy approaches could help the sector scale AI capacity while reducing resource risks and waste. 

Professor Charnley, Co-Director of the Exeter Centre for the Circular Economy, warned that the rapid growth of AI computing has exposed limitations in the traditional “take-make-dispose” model used across much of the technology industry.

“We are currently in the midst of an AI ‘gold rush,’” she said. “But this surge in demand has exposed a massive bottleneck: our linear ‘take-make-dispose’ model. We are prioritising rapid hardware replacement over long-term resilience, creating huge resource security risks.” 

The session brought together experts from across the sector to explore how circular design, procurement and operational strategies could help data centres extract more value from infrastructure already in circulation. 

Dr Diego Bermudez, Research and Impact Fellow at the Digital Innovation and Circular Economy (DICE) Network+, a UK-wide initiative led by Exeter working at the interface of digital innovation and the circular economy, highlighted the development of a Circular Data Centre Infrastructure Strategic Blueprint  

The blueprint, which includes a policy briefing with a proposed roadmap to follow, aims to help operators retain the value of hardware, critical raw materials, and energy for longer while reducing exposure to volatile supply chains. 

Panel discussions explored several practical pathways already emerging within the industry. 

Extending hardware lifecycles was identified as a major opportunity, with Astrid Wynne from Techbuyer highlighting how enterprise hardware is often retired after three to five years despite having a functional lifespan closer to eight to 10 years, creating opportunities for refurbishment and remanufacturing in secondary markets. 

The panel also discussed how modular system design and tools such as digital product passports could improve transparency around component performance and repairability, enabling “repair or refurbish by default” models across the sector. 

New operational models were also highlighted as part of the circular transition. Matt Craggs from Deep Green described how the firm is capturing waste heat from data centre operations and redirecting it to heat community facilities, demonstrating how infrastructure can create local value rather than simply consuming resources. 

Anthony Levy of Circularity First argued that procurement and financing structures will play a key role in scaling circular models. Approaches such as hardware-as-a-service could allow organisations to access computing capacity while keeping equipment within managed circular loops. 

Professor Charnley concluded that combining these approaches could help the industry respond to the growing AI demand for data centre infrastructure while avoiding future resource shortages and electronic waste challenges. 

“Data centres are designated as critical national infrastructure,” she said. “If we design systems that retain value – whether through refurbishment, modular upgrades or recovering waste heat – we can turn end-of-life assets into new value streams rather than new waste streams.” 

Throughout 2026, the DICE Network+ has been continuing its ‘deep dive’ into data centres. The Circular Centre Infrastructure Blueprint and related policy brief provide practical guidance on how to embed circular principles into the next generation of AI-driven digital ecosystems.

Find out more about the DICE Network+ here.