Microsoft, Amazon, Oracle lock in AI cloud billions

The three companies highlighted — Microsoft, Amazon, and Oracle — are not start-ups betting their existence on AI. As the article frames it, they built their fortunes on operating systems, online retail, and corporate databases, and by most normal standards they have already won. What stands out now is that each is behaving as if its AI cloud business is still on day one.

The evidence is in the contracted numbers. Microsoft began separately reporting Azure's revenue this year and revealed it is running at roughly $100 billion a year, up 43% from a year earlier, backed by a $678 billion backlog. Amazon this year locked in one of the largest cloud commitments in tech history, with Anthropic agreeing in April to spend more than $100 billion over the next decade with AWS, including up to 5 gigawatts of computing capacity built around Amazon's custom Trainium chips. Oracle's remaining performance obligations hit a record $664 billion last quarter, with roughly $300 billion tied to a single customer, OpenAI.

What the outcome hinges on is whether the demand behind these commitments holds long enough to pay off the build-out. Amazon raised its 2026 capital spending budget to about $220 billion, an enormous bet whose margins would feel it if AI demand cools before that spending pays off, and Oracle's heavy reliance on OpenAI means a stumble by that one customer could take a large chunk of future revenue with it. The article also cites Goldman Sachs chief economist Jan Hatzius, who recently repeated that AI spending cannot grow in a straight line forever — a reminder that the durability of these contracts, not their size alone, is likely to be the real test.