Meta and Microsoft have each committed approximately $50 billion in additional data center leases, pushing the combined infrastructure obligations across major cloud providers well past the $700 billion mark. When factoring in ambitious multi-year buildouts like OpenAI’s Stargate project, the total figure approaches $850 billion.

The numbers behind the AI land grab

Microsoft’s total future lease obligations now sit at roughly $155 billion. Meta’s tally comes in around $104 billion. The leases typically run 12 to 15 years and are designed to support hyperscale AI workloads, the kind of computing that powers large language models, image generators, and whatever comes next.

The aggregate commitments from leading cloud providers, including Oracle alongside Meta and Microsoft, crossed $700 billion as of March 2026. The remaining gap to $850 billion is largely attributable to OpenAI’s Stargate initiative, a sprawling multi-year data center project being built in collaboration with Oracle and other partners.

Some of these individual leases carry power capacities measured in the hundreds of megawatts. For context, a single megawatt can power roughly 750 average American homes.

The sheer scale of these obligations raises questions about off-balance-sheet risk. Lease commitments of this magnitude don’t always show up prominently in headline earnings figures, but they represent real, binding financial exposure stretched over more than a decade. If AI demand plateaus or shifts in unexpected ways, these companies would still be on the hook for billions in annual payments on facilities they may not fully utilize.

Why Big Tech is betting the farm on concrete and copper

Microsoft has OpenAI as its anchor tenant, so to speak. The company’s partnership with the ChatGPT maker means it needs to stay ahead of compute demand curves that are steepening quarter over quarter. Meta, meanwhile, is building its own family of Llama models and has made clear that AI is the strategic priority.

Oracle’s presence in this group is worth noting. The company was long considered a database legacy player, but its aggressive push into cloud infrastructure and its role in the Stargate project have repositioned it as a credible hyperscaler.

These lease commitments also create what some analysts describe as an interconnected funding loop. Tech giants sign leases, which fund data center developers, who buy chips from Nvidia and others, who then reinvest in R&D.

What this means for investors

These are 12- to 15-year bets on a technology whose commercial applications are still being defined. If enterprise AI adoption slows, or if a paradigm shift reduces compute requirements, companies could find themselves saddled with expensive, underutilized real estate.

The absence of any crypto-related components in these deals is notable for digital asset investors. Despite growing interest in decentralized compute networks and blockchain-based AI marketplaces, the actual capital flowing into AI infrastructure is coming from, and staying within, traditional corporate balance sheets. Companies like Render and Akash have pitched decentralized alternatives to centralized cloud compute, but the spending patterns here suggest Big Tech isn’t looking for partners outside its usual orbit.

Investors should watch two things closely. First, utilization rates. If these companies start reporting that newly leased capacity is filling up ahead of schedule, the bull thesis strengthens considerably. Second, power availability. Hundreds of megawatts per facility means these data centers compete with entire communities for electricity, and grid constraints in key markets could become the real bottleneck, regardless of how many leases get signed.