Amazon spent $77B+ on property and equipment in Q3, leading to a free cash outflow of $42.4B; Amazon said in July that it would likely spend $220B in 2026 capex

Amazon Web Services capacity is largely sold out for 2026, impacting your ability to secure compute resources.

Amazon reported a third-quarter free cash outflow of $42.4 billion, largely due to capital expenditures exceeding $77 billion for property and equipment. This marks a significant shift from a previous surplus. The company also increased its 2026 capital spending forecast to $220 billion, up from $200 billion, with analysts projecting further increases to over $280 billion in 2027. This heightened investment is driven by surging demand for Amazon Web Services (AWS) capacity, with CEO Andy Jassy indicating that much of the capacity for 2026 is already sold out, and even a substantial portion for 2028. The increased spending is primarily aimed at supporting the infrastructure required for artificial intelligence workloads, a sector experiencing exponential growth. This investment strategy has put pressure on cash flow, causing it to turn negative, a trend also observed in other major tech companies like Alphabet and Meta.

The substantial capital expenditure and negative free cash flow indicate Amazon's aggressive strategy to meet the immense demand for AI-driven cloud computing. This massive investment in infrastructure, with significant portions of AWS capacity already contracted for future years, signals a tightening market for compute resources. Companies relying on scalable cloud services, especially for AI development, may face increased competition for available capacity and potentially higher prices as demand outstrips supply.

This situation is not isolated to Amazon; Alphabet and Meta have also reported negative free cash flow, underscoring a broader industry trend of heavy investment in AI infrastructure. The long lead times for building out data centers and securing power mean that capacity constraints could persist for some time. Businesses needing significant AI computing power should anticipate a more challenging procurement environment and may need to explore multi-cloud strategies or long-term capacity commitments to ensure future availability.