The world's five largest technology companies—Alphabet, Meta✴, Microsoft, Amazon, and Oracle—have disclosed future commitments for leasing, facility construction, power supply, and equipment procurement totaling approximately $3.2 trillion in their latest financial reports. A significant portion of these commitments is directly tied to the massive expansion of artificial intelligence infrastructure: all five companies are striving to build extensive networks of data centers (DCs). However, the calculations also include other business areas, and some payments are spread over decades, turning the current reports into a sort of capital expenditure roadmap for the next quarter of a century.
Alphabet: Over $1 Trillion and an Eightfold Year-Over-Year Increase
The largest share of the combined commitments belongs to Alphabet. The search giant announced future expenses exceeding $1 trillion, including costs for technical equipment, electricity, and leases. According to regulatory filings, this amount is nearly eight times higher than figures from a year ago—demonstrating the pace at which the company is scaling up capital investments in computing power for AI needs. Such a leap over twelve months is unprecedented in the history of Alphabet's corporate reporting, signaling a shift from neural network experiments to industrial-scale deployment.
Meta✴: $800 Billion and a 30-Year Commitment Horizon
Meta✴ also showed a sharp surge in commitments, reporting future expenditures of over $800 billion. Just under half of this amount stems from data center leases under contracts that have not yet entered into force and will be paid over a period of up to 30 years. "We intend to continue investing aggressively to support both our core business and new AI initiatives," Meta✴ Chief Financial Officer Susan Li told analysts. Such a long commitment horizon underscores that this is not about tactical server purchases, but rather a strategic restructuring of the company's entire computing architecture for decades to come.
Negative Cash Flow and the Question of Return on Investment
Business circles continue to debate whether the hundreds of billions of dollars poured into AI server capacity will pay off. Alphabet and Meta✴—some of the most profitable companies in history—have faced negative cash flow, explaining this as a necessity to invest in a revolutionary new business direction. It is important to note that not all of these expenses are directly related to data centers: Meta✴ indicates that a portion of planned spending relates to Reality Labs consumer devices, while Alphabet and Amazon mention content licensing costs. Each company discloses information differently, making direct comparisons difficult and requiring careful interpretation of aggregated figures.
Oracle: First Negative FCF and a Plan to Return to Profitability by 2028
Oracle became the first major company to face negative free cash flow due to investments in AI data centers. At an investor presentation, the company outlined a plan to return to positive territory by 2028. According to management, supplier financing and customer prepayments will help reduce the need for internal funds in the early stages of projects. This approach—raising external capital to cover capital expenditures—could serve as a model for other players in the race for AI capacity, especially those whose cash reserves do not match Alphabet or Microsoft.
Methodology and Caveats: Why $3.2 Trillion Is a Lower Bound
Combined, the five companies' future lease commitments stand at $1.3 trillion, with other procurement obligations totaling $1.9 trillion. All these figures were provided by the companies themselves in their latest financial reports, except for Microsoft, which disclosed its procurement and construction order data in an annual report covering the period through the end of June. Significant expense items, including a portion of lease and debt obligations, are not included in this analysis. Thus, $3.2 trillion represents the lower bound of the tech sector's actual future infrastructure costs rather than its total volume, and as companies refine their reporting, the final figure is likely to grow.
