Alphabet, Meta Platforms, Microsoft and Amazon have committed nearly $2.4 trillion to future spending, signaling that investment in artificial intelligence infrastructure will continue to rise sharply over the coming years.

As the companies race to expand their AI computing capacity, their commitments related to data center leases, construction, energy supplies and technology equipment have increased substantially over the past year. The obligations include both near-term purchases and long-term agreements extending for several decades.

Google parent Alphabet disclosed last week that its purchase commitments, contractual obligations and leases that have not yet begun totaled $902 billion—more than nine times the amount reported a year earlier. Regulatory filings show that the commitments cover technology equipment, energy supplies and leasing agreements.

Meta reported a similarly sharp increase, with future spending commitments approaching $700 billion, more than eight times the previous year’s total. About half of that amount relates to data center leases that have not yet commenced, with some payments extending for as long as 30 years.

The figures do not represent immediate capital expenditures alone. They also include long-term leasing, purchasing and energy obligations that will be paid over many years.

The rapid expansion has intensified debate over whether the hundreds of billions of dollars being invested in AI servers and computing clusters will generate sufficient revenue and returns. Free cash flow at Alphabet and Amazon has already turned negative, while Meta is expected to face similar pressure.

Despite growing concerns about cash flow, the companies indicated in their latest earnings reports that spending will continue to increase. They cited surging demand for computing capacity driven by generative AI, cloud services and enterprise applications.