The Multi-Billion Dollar Race to Control AI's Physical Infrastructure

2026-03-19 companies

San Francisco, Wednesday, 18 March 2026.
Tech giants are aggressively securing power and data centers to dominate the AI market, highlighted by a projected $450 billion infrastructure spend in 2026 to build competitive moats.

The Mega-Deals Defining the AI Era

In a defining move for the artificial intelligence sector, Meta Platforms Inc. (META) announced on Monday, March 16, 2026, a sweeping long-term agreement to secure AI infrastructure from the Dutch cloud provider Nebius Group NV (NBIS) [1][2]. The transaction, valued at up to $27 billion over a five-year period, underscores a critical pivot in the tech industry: the race for software supremacy is now inextricably linked to the control of physical hardware [1]. Under the terms of the deal, Nebius will provide Meta with $12 billion in dedicated capacity beginning in early 2027, with Meta committing to potentially purchase an additional $15 billion in available compute capacity [1][2]. The market reacted swiftly to the news, sending Nebius shares surging 14% in early trading [1].

Securing the Foundations of Compute

The scale of capital expenditure required to maintain a competitive moat in the AI space has reached unprecedented levels [GPT]. Meta alone projects its AI-related capital expenditures will reach between $115 billion and $135 billion this year [1]. Broader estimates suggest that the “big five” hyperscalers—Amazon, Meta, Alphabet, Oracle, and Microsoft—are projected to allocate upwards of $450 billion in capital expenditure specifically for AI infrastructure in 2026 [4]. Some industry analyses project total combined spending by hyperscalers could reach an astonishing $700 billion [1]. Even pure-play AI research firms are transitioning into infrastructure managers; OpenAI, for instance, is currently expanding its “Stargate” AI rack infrastructure in San Francisco, actively recruiting engineers to design, manufacture, and deploy proprietary datacenter systems across a global network [5].

Real Estate Investment Trusts Capitalize on Demand

The physical footprint required for AI training and inference has also catalyzed immense growth within data center Real Estate Investment Trusts (REITs) [4]. These specialized trusts own and manage the customized facilities that house IT infrastructure, providing the critical power and cooling necessary for advanced compute [4]. Equinix (EQIX), a prominent data center REIT, recently signaled its aggressive expansion strategy by agreeing in early March 2026 to acquire the Nordic data center operator atNorth for $4 billion, in partnership with the Canada Pension Plan Investment Board [4]. Looking ahead, Equinix plans to invest between $4 billion and $5 billion annually from 2026 through 2029 [alert! ‘Forward-looking projections are subject to market conditions and regulatory approval’], with the goal of doubling its data center capacity [4].

Sources

Artificial intelligence Infrastructure