Data Centre REITs Outperform S&P 500 for Third Consecutive Year

Data centre REITs have outperformed the S&P 500 for the third consecutive year, delivering 28% average total returns versus the broader market's 15% gain. This three-year run of outperformance - cumulative returns exceeding 90% since early 2024 - has made data centre REITs the best-performing real estate subsector and one of the best-performing segments of the entire equity market. The performance reflects a fundamental rerating of the sector driven by AI-related demand that shows no signs of abating.

Equinix, with a market capitalisation now exceeding $80 billion, has become the most valuable REIT of any property type globally - surpassing logistics giant Prologis and cell tower operator American Tower. The company reported 12% year-over-year revenue growth with particularly strong interconnection metrics. Its network of 270+ data centres across 75 metropolitan areas creates an ecosystem effect where each additional customer makes the platform more valuable for existing tenants, providing a competitive moat that is extremely difficult to replicate.

Digital Realty saw its largest-ever leasing quarter at 180 MW of new signings. The company's strategic pivot toward hyperscale under CEO Andy Power has been vindicated by the AI infrastructure surge, which favours large-scale deployments that align with Digital Realty's campus development model. CyrusOne, QTS (within Blackstone's portfolio), and Vantage have similarly reported record backlog levels across their respective portfolios.

The private equity market has been even more active than the public markets. Blackstone, KKR, Brookfield, Apollo, and DigitalBridge have collectively deployed over $50 billion in data centre assets over the past three years. Blackstone alone holds what may be the largest private data centre portfolio globally through its QTS platform and numerous other investments. The firm is reportedly considering a $2 billion IPO for its data centre unit, which would create another major publicly traded operator. The $40 billion Aligned Data Centers acquisition - involving Microsoft, NVIDIA, xAI, and several sovereign wealth funds - set a new record for the largest infrastructure transaction of any type.

The investment thesis rests on several structural advantages that distinguish data centres from other real estate sectors. First, long-term contractual revenues: the average data centre lease term is 8-12 years with built-in annual escalators of 2-4%, providing predictable, inflation-protected cashflows. Second, high switching costs: moving out of a data centre - migrating servers, re-establishing network connections, and requalifying redundancy systems - typically costs tenants $5,000-15,000 per kW, creating substantial lock-in. Third, the barriers to entry created by power infrastructure, fibre connectivity, and increasingly, regulatory approvals, protect existing operators from new competition.

The bear case centres on valuation. Equinix trades at approximately 28-30x forward funds from operations (FFO), compared to the traditional REIT average of 15-18x. Digital Realty trades at 24-26x. These premium multiples assume sustained double-digit demand growth driven by AI infrastructure, but if AI adoption follows a traditional S-curve rather than the linear trajectory currently priced in, the sector could face overcapacity in the late 2020s. Additionally, the massive construction pipeline - estimated at $121 billion annually - will eventually deliver substantial new supply that could compress occupancy rates and lease pricing.

For institutional investors, the calculus remains compelling. Data centres offer a rare combination of growth and defensiveness in the real estate portfolio: the infrastructure-like characteristics provide downside protection through long-term contracts with investment-grade tenants, while the AI demand driver provides exposure to the fastest-growing segment of the global economy. The key risk is paying too much - at current valuations, investors are pricing in a very specific view of AI's trajectory that leaves little margin for error.