Amazon CEO Andy Jassy said Amazon Web Services has booked most of its cloud-computing capacity through 2027, with sizable commitments already extending into 2028, as enterprise demand for AI infrastructure continues to outpace supply1,2.
Speaking on a video podcast hosted by Andreessen Horowitz's Ben Horowitz and Erik Torenberg, Jassy described demand for AI infrastructure as "insatiable" and said even AWS capacity for 2026 is effectively sold out. The disclosure reframes the competitive picture for hyperscale cloud providers: the constraint is not customer demand but how fast Amazon can build data centers and procure advanced chips to deliver what has already been reserved.
Amazon has pointed to roughly $220 billion of capital expenditure for 2026, and Jassy suggested the company could spend more, implying a long runway of heavy investment before the corresponding revenue fully lands. AWS does retain some compute for on-demand users and for services such as Bedrock, its managed AI-model access layer, and EC2 virtual servers, but Jassy said the large, longer-term capacity blocks are largely committed.
ANALYSIS The timeline Jassy outlined means AWS is now selling infrastructure two to three years ahead of delivery, a cadence more typical of power-plant or semiconductor-fab procurement than traditional cloud computing. That forward-sold posture raises execution risk: any delay in data-center construction or chip supply would push committed capacity further out.