Calls from major artificial intelligence companies to slow the pace of AI development have rattled related stocks in recent days, with the impact extending beyond chipmakers and technology firms to one of the biggest physical infrastructure bets tied to the AI boom: data centres.The real estate sector has significant exposure to the expansion. While cloud computing, storage, enterprise IT and internet services all require data centre capacity, AI has rapidly emerged as the dominant source of demand.AI could account for about 70% of global data centre capacity demand by 2030, according to McKinsey. The consultancy estimates that meeting total global data centre demand by then will require nearly $7 trillion in capital spending.The real estate component alone could account for about $3 trillion of investment over the next five years, according to JLL.The concerns were reflected in markets this week, when Digital Realty and Equinix, two of the largest data centre real estate investment trusts (REITs), saw their stocks fall following warnings from AI leaders over the pace of development.
Data centre demand extends beyond AI
Digital Realty CEO Andrew Power said the calls for slower AI development by Anthropic, OpenAI and xAI did not mean work on the infrastructure supporting the broader digital economy would stop.“There’s tremendous digital transformation happening that is not connected to AI,” Power told CNBC’s Property Play.“There is tremendous cloud computing growth. Frankly, from my business lens, my seat, I think those demand trends, which are massive drivers of our business, have been stifled in these days of AI.”Power said hyperscalers had been forced to choose between expanding their commercial cloud businesses and allocating capacity to AI laboratories.Digital Realty operates in markets including Northern Virginia, Dallas, Chicago, Singapore, Tokyo, Frankfurt and Amsterdam. Power said customers in these locations were competing for the same data centre space.“Our markets’ demand has been outpacing supply now for several years. There’s pent-up need for infrastructure in those markets. There’s locational sensitivity. Those workloads can’t choose any one of the 50 states,” he said.The company also has a global portfolio, which Power said helps meet requirements related to data sovereignty and customer support in different countries.
AI inference could keep demand growing
JLL’s global head of data centre research and strategy, Andrew Batson, said a slowdown in the development of new AI models would not necessarily translate into an equivalent slowdown in physical data centre requirements.“The real growth in data centers over the next handful of years is in inference — that’s the adoption by businesses and citizens of the tool into daily workflow,” Batson said.He noted that only one in four Americans currently use AI daily, leaving significant room for adoption to increase.“So even if models are slow to be released, there is significant runway for adoption to grow and data center demand to increase,” he said.Batson also pointed to institutional investors including Blackstone, BlackRock and KKR, saying they “have high conviction in this space”.“That, on paper, still looks quite strong, despite some of the headlines here,” he added.
Digital Realty’s $20 billion pipeline
Digital Realty’s development pipeline currently totals $20 billion under construction, according to the company, compared with $10 billion at the end of 2023.Power said the company had also changed how it funds its capital-intensive business, including raising private capital and using individual joint ventures.“The first, most important part is, make sure that the daily gyrations, our stock price, don’t affect our strategy, our business,” Power said.“We evolved our funding model a couple of years ago. We are an incredibly capital-intensive business.”He said Digital Realty had positioned its balance sheet with high liquidity and lower leverage.“We’re on to the next iteration of that in raising private capital. We’ve also done one-off joint ventures, and we positioned the balance sheet in probably the most liquidity, the lowest leverage, the best place it could be in any potential storm,” Power said.He added that he was “not suggesting today is an end-of-the-world storm or anything like that”.For data centre real estate, the debate over the pace of AI development therefore comes alongside a broader demand picture involving cloud computing, enterprise technology and the still-expanding adoption of AI tools.