Data centre REITs ride AI-driven boom amid underlying fragilities

Data centre real estate investment trusts (REITs) are enjoying a boom as demand for AI-ready infrastructure surges, but analysts caution the sector’s rapid rise may rest on fragile financial and operational foundations.

Fitch Ratings upgraded its outlook for data centre REITs to “improving” in a sector assessment dated 9 December 2025, one of only two REIT subsectors to receive a positive revision. “Data center and office REITs are facing strong tenant demand, with the former benefiting from AI-driven investment and the latter supported by RTO initiatives favoring the most desirable workspace,” Christopher Wimmer, senior director at Fitch Ratings, said in the report.

Market activity underlines the momentum. According to ITPro, global data centre investment reached a record $61 billion in 2025, surpassing 2024, driven by hyperscaler expansion, private equity interest and higher debt financing. ITPro cites planned capacity increases that include 114 hyperscaler data centres already planned or under construction and notes major initiatives such as Meta’s partnership with Blue Owl Capital and a multibillion-dollar fund announced by Microsoft and MGX Fund Management. Industry consultancy McKinsey is cited in that coverage projecting capital needs for global data centres of as much as $6.7 trillion by 2030, with roughly $5.2 trillion tied to AI computing demands.

Performance metrics in public markets have reflected investor confidence. Research compiled by Ginlix.ai shows the weighted average AFFO multiple for data centre REITs at 23x in July 2025, above the 21x average for the broader REIT market. Cleanbridge’s market report highlights compressed REIT implied cap rates for data centres, which fell to 4.23% in the fourth quarter of 2024, well below private market cap rates, indicating an enthusiastic re-rating of listed assets.

Yet the tailwinds mask material risks. Several industry reports and company updates point to structural and monetisation challenges. Keppel DC REIT’s 3Q 2025 operational presentation emphasises a sharp rise in AI workloads in Asia-Pacific, with AI’s share of demand in that region rising from 4.9% in 3Q 2024 to 8.8% in 3Q 2025 and forecasts that global data centre demand could reach 100 GW by 2027. Mapletree Industrial Trust’s mid‑2025 analysis expects AI capacity demand to grow at roughly 33% a year between 2023 and 2030, and projects advanced‑AI workloads will make up about 70% of total demand by 2030, with generative AI accounting for 40%.

Those bullish demand projections, however, collide with questions over who ultimately pays and whether hyperscalers and tech firms can convert massive capital spending into sustainable returns. Industry commentary assembled by Ginlix.ai and market observers warns that hyperscalers may be expanding capacity to secure strategic advantage rather than to chase immediate margin accretion, leaving REITs exposed to tenant concentration, concessionary pricing and the long lead times required to turn capacity into revenue. Minichart’s market analysis highlights practical constraints: the exponential energy and water needs of large AI training operations, the growing scale of hyperscale facilities (20–50MW+), and the difficulty of retrofitting older sites given utility and lease limitations.

The evolving nature of AI workloads adds further complexity. Mapletree’s report points to a likely structural shift by 2028, when inference workloads, demanding low‑latency, distributed edge capacity, may account for 80% of AI demand versus 20% for training. That transition will favour edge and regional facilities over a small number of gigawatt‑scale training sites, potentially reshaping leasing patterns and capital allocation across the sector.

Financing dynamics deepen the tension. Record investment flows and compressed public market yields suggest ample capital is available, but Cleanbridge’s comparison of public versus private market cap rates shows a divergence that could reverse if investor sentiment changes. Higher AFFO multiples in the listed REIT space have priced in future growth; any slowdown in hyperscaler spending or a reassessment of AI ROI would expose valuations to rapid re-pricing.

REIT managers and developers are responding with a mix of strategies. Keppel and other listed operators are prioritising scalable, resilient designs and targeting long‑term contracts with diversified customer rosters to reduce concentration risk. ITPro’s coverage notes increased use of joint ventures and financial sponsors to spread development risk and finance the large, multi‑year projects required by AI workloads. At the same time, operators are wrestling with the environmental and grid implications of gigawatt‑scale deployments, including the need for dedicated power plants, upgraded substations and water management, constraints that can slow delivery and raise capex intensity.

For investors the picture is therefore nuanced. Public market metrics and record capital commitments signal confidence in data centre REITs as a secular play on digital transformation and AI. Yet multiple independent analyses and company disclosures underline that the sector’s growth depends on continued hyperscaler commitment, successful monetisation of new capacity, and navigation of energy, water and logistical bottlenecks. As one industry observer summarised, the sector is benefitting from a generational tailwind, but the commercial and operational economics underpinning that tailwind remain uneven and must be proven over time.