© David Ryder / Getty Images
I keep buying Amazon (NASDAQ:AMZN | AMZN Price Prediction), and the July 30, 2026 earnings report gave me another reason to keep the buy button warm. When a company at a $2.921 trillion market cap accelerates its cloud business to its fastest pace in 18 quarters while committing to what looks like more than $220 billion in capital spending this year, I read that as a demand problem I want to own.
The thesis in plain English: Amazon is building 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 carries a $496B contracted backlog, which is the receipt for the spending.
Three Reasons the Conviction Holds
First, AWS growth. Segment revenue reached $42.23 billion in Q2 2026, up 37% year over year, at a 39.4% operating margin. Andy Jassy told investors that Amazon’s “AI and Chips businesses each eclipsed run rates of more than $25 billion.” That is a booked run rate compounding at triple-digit percentages.
Second, the earnings power funding the buildout. Q2 operating income rose 43.24% to $27.461 billion. EPS came in at $5.75 against a $1.8227 estimate, a 215.47% beat. Full-year 2025 operating cash flow reached $139.514 billion. The balance sheet absorbs the strain: net debt to EBITDA of 0.45, interest coverage of 35.17x, and shareholders’ equity at $551.62 billion.
Third, the capex trajectory. H1 2026 capex hit $98.411 billion. Q2 alone was $54.208 billion, up 68.44% year over year. Polymarket traders assign a 62.5% probability that 2026 capex clears $220 billion. A hyperscaler outrunning its own capex guide because customers keep asking for more is the signal I want to see.
Why Amazon and Not the Obvious Alternatives
A reader might reach first for Microsoft (NASDAQ:MSFT) or Alphabet (NASDAQ:GOOGL). Both are quality operators. My money keeps going to Amazon because the customer pull evidence sits here in signed form. Project Rainier puts more than 500,000 Trainium2 chips behind Anthropic’s Claude training. Two of the most compute-hungry AI labs on the planet are running production workloads on silicon Amazon designed in-house, at a 39.4% AWS operating margin. Capacity is sold out through 2027 with booked commitments extending into 2028. That is a moat I can measure in contracts.
The Real Risk
Free cash flow 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. If AI demand softens before this capacity monetizes, the return on invested capital math turns ugly. I have made peace with it because the demand is contractual, and the Q3 guide of $22.5 billion to $26.5 billion in operating income shows the operating engine compounding through the buildout.
What Keeps the Buy Button Active
Amazon has told me it wants to spend more than $220 billion this year because customers made it. AWS growing 37% year over year at scale, an AI silicon franchise the two most demanding labs on earth already run on, and a Q3 operating income guide that dwarfs the year-ago quarter tells me this capex cycle is a wealth-building event I want to be part of. That is exactly why my next buy is already scheduled.
Contact [email protected] for any questions or corrections.
