Between July and September 2026, Alphabet, Microsoft and Amazon spent a combined $168 billion on property and equipment, according to the quarterly results they published on 28 and 29 October. A year earlier, the same three companies spent about $78 billion in the same quarter. Most of this money goes into data centres, servers and chips, and all three companies tie their growth to demand for AI.
The numbers, company by company
These are the cash figures each company reported for purchases of property and equipment in the quarter:
- Amazon spent $77.3 billion, up from $35.1 billion a year earlier.
- Alphabet spent $51.4 billion, up from $24.0 billion.
- Microsoft spent $39.4 billion, up from $19.4 billion. That figure leaves out finance leases, so Microsoft's full spending on data centres is higher.
Each company roughly doubled its spending in twelve months. Amazon also said it still expects to spend about $220 billion on capital in 2026.
The demand side is growing fast too
The spending lines up with fast growth in cloud revenue, which is where companies rent computing power, including for AI models. Google Cloud revenue grew 103% to $30.7 billion, and Alphabet said the growth was led by enterprise AI infrastructure and AI solutions. Microsoft said Azure revenue grew 48%. Amazon Web Services grew 39% to about $46 billion, which Amazon called its fastest growth in nearly five years.
Microsoft's results also showed how valuable a single AI customer can be. TechCrunch reports that Microsoft's quarterly filing shows about $10 billion of revenue from its commercial arrangements with OpenAI, nearly 11% of Microsoft's $92.3 billion total for the quarter.
The bills are showing up in cash flow
Building at this pace costs more than these businesses currently bring in after expenses. Amazon reported free cash flow of minus $42.4 billion over the last twelve months and said that mainly reflects its AI investments. Alphabet reported free cash flow of minus $9.1 billion for the quarter. Both companies remain very profitable, and these figures show how heavily they are betting that AI demand keeps growing. Amazon also missed analyst expectations on revenue and earnings per share, according to TechCrunch, while AWS beat them.
A useful historical comparison
There is an honest parallel in the late 1990s, when telecom companies laid huge amounts of fibre-optic cable in expectation of internet demand. Demand arrived more slowly than planned and many of those companies went bust, but the cheap capacity they left behind carried the internet for the next two decades. Today's spenders are far more profitable than most of those telecoms were, so the comparison is about the pattern of building ahead of demand, and the likely benefit to users once the capacity exists.
What to do this week
For most people the practical effect is that AI tools should keep getting more capable and cheaper to run, while vendors push harder for daily use to justify the spending. This week, list every AI tool your team pays for and note whether each one charges per seat or per use. Pricing is likely to keep changing as these bills come due, and knowing your mix makes it easier to negotiate or switch.
For more on Alphabet's quarter, read Gemini reaching one billion users, and see Stacks for the tool combinations teams use most.