Amazon commits to $220+ billion in AI capex for 2026, highest single-year infrastructure spend by any hyperscaler.
Amazon keeps finding reasons to expand its artificial intelligence footprint, and its July 30, 2026 earnings report provided yet another compelling justification. With a market capitalization of $2.921 trillion, the company accelerated its cloud business to its fastest pace in eighteen quarters while committing to capital spending exceeding $220 billion this year. When a company of this scale demonstrates such aggressive infrastructure buildout, it signals a demand problem worth owning.
The underlying thesis is straightforward: Amazon is constructing AI capacity to fulfill contracts customers have already signed. Anthropic committed up to 5 GW of Trainium chip capacity, OpenAI committed roughly 2 GW of Trainium capacity beginning 2027, and Amazon Bedrock customers spent more in Q2 than all prior quarters combined. AWS maintains a $496 billion contracted backlog, which underwrites the spending trajectory.
AWS segment revenue reached $42.23 billion in Q2 2026, a 37% year-over-year increase, generating a 39.4% operating margin. CEO Andy Jassy told investors that Amazon's AI and Chips businesses each eclipsed run rates of more than $25 billion, representing a booked run rate compounding at triple-digit percentages. Q2 operating income rose 43.24% to $27.461 billion, while earnings per share came in at $5.75 against a $1.8227 estimate—a 215.47% beat. Full-year 2025 operating cash flow reached $139.514 billion, and the balance sheet handles the strain comfortably: net debt to EBITDA of 0.45, interest coverage of 35.17x, and shareholders' equity at $551.62 billion.
Through the first half of 2026, capex hit $98.411 billion, with Q2 alone accounting for $54.208 billion, a 68.44% year-over-year increase. Polymarket traders assign a 62.5% probability that 2026 capex clears $220 billion. A hyperscaler outrunning its own capex guidance because customers keep requesting more capacity represents precisely the signal worth watching.
While investors might gravitate toward Microsoft or Alphabet—both quality operators—the compelling case for Amazon rests on the visible, contractual evidence of customer demand. Project Rainier deploys more than 500,000 Trainium2 chips behind Anthropic's Claude training. Two of the most compute-intensive AI laboratories on the planet run production workloads on silicon Amazon designed internally, achieving a 39.4% AWS operating margin. Capacity is sold out through 2027 with committed bookings extending into 2028—a defensible competitive position measured in signed contracts.
The obvious risk deserves acknowledgment. Free cash flow over the trailing twelve months turned negative at -$7.6 billion because capex now consumes 138% of first-half operating cash flow. Long-term debt climbed to $119.1 billion from $65.6 billion. Should AI demand soften before this capacity monetizes, the return-on-invested-capital calculus deteriorates. However, the demand is contractual rather than speculative, and the Q3 operating income guide of $22.5 billion to $26.5 billion demonstrates the operating engine compounding through the buildout cycle.
Amazon is spending more than $220 billion this year because customers made it necessary. AWS growing 37% year-over-year at scale, an AI silicon franchise that the two most demanding laboratories on earth already depend on, and a Q3 operating income guide that substantially exceeds the year-ago quarter all signal that this capex cycle represents a wealth-building event worth participating in.