Big Tech spent $165 billion on AI buildout in one quarter

Capital expenditure is the money a company spends on long-lived assets: data centers, servers, chips, networking, land and power equipment. For the cloud giants it has become the clearest gauge of how hard they are pushing on AI, because almost all of the increase goes to AI computing capacity.

The numbers from the filings

We took each company's line for purchases of property and equipment from the cash flow statements in its quarterly filings and earnings releases. Microsoft's fiscal year ends in June, so its April to June quarter is its fiscal fourth quarter; we derived it by subtracting nine-month figures from full-year figures in its 10-K.

Combined, the four spent $165.1 billion in the quarter, against $88.2 billion in April to June 2025. That is an increase of $76.8 billion, or about 87%, in one year nContext: n=4 companies · Apr to Jun 2026 vs Apr to Jun 2025 · company cash flow statements, SEC filings · margin: not applicable, reported figures.

What each company said about AI

• Amazon spent $54.2 billion. Its release says the year-over-year jump in trailing twelve-month capital spending, $66.1 billion, 'primarily reflects investments in artificial intelligence.' AWS revenue rose 37% to $42.2 billion from $30.9 billion nContext: n=1 segment (AWS) · Q2 2026 vs Q2 2025 · Amazon Form 8-K, Exhibit 99.1 · margin: not applicable, reported figures, and Amazon said its AWS AI business and its chips business each passed a $25 billion annual revenue run rate.
• Alphabet spent $44.9 billion. Google Cloud revenue grew to $24.8 billion from $13.6 billion a year earlier, which the company attributed to demand for AI infrastructure and AI solutions. Alphabet also raised $49.6 billion in equity in June, earmarked in part for 'capital expenditures to scale AI infrastructure and global compute.'
• Microsoft spent $35.8 billion in the quarter and $115.9 billion over its full fiscal year 2026, against $64.6 billion in fiscal 2025. Azure revenue passed $100 billion for the year, according to CEO Satya Nadella, and the company reported a commercial remaining performance obligation of $678 billion.
• Meta spent $30.1 billion on property and equipment, or $31.1 billion including finance-lease payments. It narrowed its 2026 capital spending outlook to $130 billion to $145 billion, from $125 billion to $145 billion.

Definitions differ slightly. Microsoft's figure is cash additions to property and equipment and excludes finance leases; Meta also reports $1.0 billion of finance-lease principal payments on top of its $30.1 billion. We use each company's main cash line so the four numbers are comparable to what each reports as capital expenditures.

Revenue is growing, but slower than spending

Cloud revenue is growing fast at all three cloud providers. But spending is growing faster. Alphabet's capital spending doubled, to $44.9 billion from $22.4 billion, while its total revenue grew to $119.8 billion from $96.4 billion. Meta's capital spending rose to $30.1 billion from $16.5 billion, while revenue rose to $60.8 billion from $47.5 billion. Spending now consumes most of the cash these companies generate, a point we examine in a separate piece on the AI bubble question.

Why these numbers matter for the rest of the market

One company's capital spending is another company's revenue. NVIDIA reported $48.7 billion of revenue from hyperscale customers in its quarter ended July 26, 2026, according to its CFO commentary. Memory makers, equipment makers, power suppliers and data-center builders are all downstream of the same budgets. When these four companies report third-quarter results in late October, their capex lines will be the first numbers many AI traders look at.