Alphabet just told investors it has signed $811 billion in long-term purchase commitments, mostly for artificial intelligence infrastructure, a figure that dwarfs the $332 billion on its books three months earlier and signals the company expects the current compute shortage to persist for years. The disclosure arrived alongside a second quarter in which the company posted negative free cash flow for the first time since its 2004 initial public offering, after spending $45 billion on capital expenditures, double the pace of a year ago. Management also raised its full-year 2026 capex budget to between $195 billion and $205 billion and warned that spending will increase significantly in 2027.
The contracts run deep
The commitments are not balance-sheet liabilities, but they are binding take-or-pay agreements that lock in chip supply, data-center construction and energy services. Alphabet said it expects to fulfill the bulk of them by 2030, while energy contracts stretch as far as 2054. That timeline makes clear the company is not building for a hype cycle; it is securing the physical layer of a business it believes will still be expanding decades from now. The $520 billion in remaining performance obligations at the end of June, essentially contracted future revenue, suggests customers are signing up for capacity that does not yet exist.
The bridge is expensive
A severe shortage of compute capacity has forced Alphabet to rent third-party infrastructure as a bridge until its own data centers come online, a decision that will compress margins in the near term. Inventory jumped from $2.4 billion to $10 billion last quarter, partly reflecting the ramp of custom Tensor Processing Unit systems the company now sells directly. Those TPU sales require their own long-lead commitments to chip-design partners, feeding the same $811 billion pipeline. The logic is circular: more capacity enables more sales, which justify more capacity.
The cash cows still moo
Investors who flinch at the headline number might note that the core advertising business and the cloud segment continue to generate substantial cash and, in the cloud’s case, strong returns on invested capital. The negative free cash flow is a choice, not a crisis, a deliberate decision to front-load spending on assets with 26-year energy contracts. Whether the demand curve bends before 2030 is the only question that matters. Alphabet has bet $811 billion that it will not.