Contrary to all the gloomy forecasts about an artificial intelligence bubble, the numbers speak for themselves. In the second quarter of 2026, the combined cloud revenue of Amazon, Google, and Microsoft jumped 48% year-over-year. This is an impressive acceleration compared to 39% a quarter earlier and contrasts particularly sharply with the modest 24–33% seen in 2024–2025. For companies with turnover in the hundreds of billions of dollars, such momentum is a true anomaly, breaking the usual logic of growth slowing as scale increases.
Numbers That Cannot Be Ignored
A breakdown of the financial reports shows that skeptics doubting the return on AI investments clearly jumped to conclusions too quickly. AWS posted revenue of $42.2 billion, up 37% from a year earlier — its fastest growth in 18 quarters. The division's operating margin stood at an impressive 36.8%.
Google Cloud delivered even more striking results: revenue reached $24.8 billion, up 82%. Operating profit more than tripled over the year, while the margin rose from 20.7% to 35.6%. Azure, for its part, crossed the symbolic milestone of $100 billion in annual revenue, showing 43% growth in constant currency.
Market forecasts now paint an even rosier picture: the combined revenue of Amazon, Google, Meta, and Microsoft is set to grow from $1.6 trillion in 2024 to $2.5 trillion within two years. Average annual growth rates relative to revenue over the last 12 months have jumped from 11% to 18%.
Skeptics' Arguments and Oracle's Position
Critics, however, point to a worrying signal: according to PIMCO estimates, capital expenditures by the largest cloud providers will absorb about 94% of their operating cash flow over the next two years. By comparison, this figure did not exceed 40% in 2023. Fidelity, on the other hand, sees a different picture: the ratio of capital expenditures to free cash flow across the broad market is currently below one, whereas at the peak of the dot-com era in 2000 it approached four.
Oracle deserves special attention. The company's outstanding contractual obligations have reached $638 billion — a 363% increase year-over-year. These contracts are already signed, although the revenue from them has not yet been recognized in the financial statements. When the data centers reach full capacity, we will see the same profitability growth scenario, but with a time lag.
My verdict: the infrastructure layer of AI has already proven its viability, monetizing computing and data storage. However, the key question — how the corporate sector will profit from the AI applications themselves — remains open. This is where both the main potential and the main risk for the market lie.