AI stocks and the data center buildout
The largest cloud companies are spending at a record pace on data centers, chips and power to meet demand for AI. These charts track that spending from company filings, how much of their cash flow it uses, and how the stocks supplying the buildout have performed.
Updated- Hyperscaler capex, last 4 quarters
- $566B
- +81% from a year earlier
- Capex as a share of operating cash flow
- 82%
- 61% a year earlier
- Free cash flow after capex
- $126B
- -38% from a year earlier
- AI stocks' share of the S&P 500
- 36%
- 45 of 503 stocks, by market value
Hyperscalers here are Microsoft, Amazon, Alphabet, Meta and Oracle. Combined figures run through Q2 2026, the latest quarter all five have reported; Meta and Oracle have also reported Q3 2026.
Hyperscaler capex by quarter
Cash spent on property and equipment each calendar quarter: mostly servers, chips, data center buildings and the land and power to run them.
Source: company 10-Q and 10-K filings via SEC EDGAR
Hyperscaler capex by year
The latest year includes the quarters reported so far.
Source: company 10-Q and 10-K filings via SEC EDGAR
Capex vs operating cash flow
Trailing four quarters, all five companies combined. When capex approaches operating cash flow, less is left for buybacks, dividends and cash on the balance sheet.
Source: company 10-Q and 10-K filings via SEC EDGAR
Free cash flow after capex
Operating cash flow minus capex over the last four quarters, with the year-earlier figure underneath.
- Microsoft$67.0B$71.6B a year ago
- Alphabet$53.3B$66.7B a year ago
- Meta$41.0B$50.1B a year ago
- Amazon-$11.6B$13.5B a year ago
- Oracle-$23.7B-$394.0M a year ago
Source: company 10-Q and 10-K filings via SEC EDGAR
Nvidia's revenue vs hyperscaler capex
Quarterly. A large share of hyperscaler capex goes to AI chips and the systems built around them, and Nvidia is the biggest seller. Nvidia's fiscal quarters end about a month after calendar quarters.
Source: company 10-Q and 10-K filings via SEC EDGAR
Capex as a share of revenue
Trailing four quarters. A rising line means each dollar of sales is funding more construction.
Source: company 10-Q and 10-K filings via SEC EDGAR
The AI stack
45 S&P 500 companies grouped by the part of the buildout they supply, worth $26.8T together, or 36% of the index.
Year-to-date return by layer
Weighted by market value.
- Servers+228%
- Memory and storage+217%
- Chip equipment+37%
- Networking and optics+37%
- Chips+33%
- Data center REITs+24%
- Power and cooling+11%
- Hyperscalers+4%
- Software+2%
Source: FI Handbook, S&P 500 members by market value
Hyperscalers$12.4T · +4% this year
The cloud and AI platforms writing the checks.
| Company | Market cap | Today | YTD | 1 year | P/E |
|---|---|---|---|---|---|
| $4T | -0.61% | +3% | +27% | 16 | |
| $4T | -0.31% | +9% | +2% | 29 | |
| $3T | -0.06% | +7% | +12% | 20 | |
| $2T | -1.46% | -2% | -12% | 22 | |
| $378B | -1.14% | -34% | -54% | 20 |
Chips$8.4T · +33% this year
GPUs, custom accelerators and the CPUs around them.
| Company | Market cap | Today | YTD | 1 year | P/E |
|---|---|---|---|---|---|
| $5T | +1.80% | +13% | +16% | 27 | |
| $2T | +1.12% | -7% | -6% | 42 | |
| $855B | +1.33% | +141% | +120% | 134 | |
| $462B | +0.47% | +136% | +128% | — | |
| $206B | +0.25% | +169% | +177% | 75 | |
| $188B | +0.03% | +4% | +6% | 20 |
Chip equipment$841.8B · +37% this year
The tools that make chips: deposition, etch and inspection.
| Company | Market cap | Today | YTD | 1 year | P/E |
|---|---|---|---|---|---|
| $304B | +1.61% | +47% | +66% | 33 | |
| $295B | +2.38% | +35% | +57% | 35 | |
| $194B | +2.08% | +20% | +26% | 41 | |
| $49B | +1.00% | +60% | +113% | 43 |
Memory and storage$1.5T · +217% this year
High-bandwidth memory, flash and hard drives.
| Company | Market cap | Today | YTD | 1 year | P/E |
|---|---|---|---|---|---|
| $1T | +3.30% | +208% | +325% | 12 | |
| $184B | +2.82% | +422% | +641% | 17 | |
| $146B | +1.54% | +130% | +180% | 46 | |
| $131B | +1.67% | +100% | +174% | 12 |
Networking and optics$1.1T · +37% this year
Switches, optical links, connectors and fiber.
| Company | Market cap | Today | YTD | 1 year | P/E |
|---|---|---|---|---|---|
| $428B | +1.82% | +41% | +54% | 33 | |
| $220B | +1.03% | +32% | +13% | 55 | |
| $183B | -0.01% | +10% | +10% | 37 | |
| $107B | +0.18% | +42% | +45% | 57 | |
| $71B | +4.28% | +105% | +349% | — | |
| $46B | +1.22% | +37% | +86% | 73 | |
| $42B | +2.07% | +13% | +72% | 52 |
Servers$393.3B · +228% this year
The racks the chips go into.
| Company | Market cap | Today | YTD | 1 year | P/E |
|---|---|---|---|---|---|
| $296B | +1.98% | +266% | +199% | 27 | |
| $76B | +0.67% | +139% | +149% | 29 | |
| $22B | -0.24% | +13% | -31% | 10 |
Power and cooling$950.8B · +11% this year
Turbines, electrical gear, cooling and the utilities selling the power.
| Company | Market cap | Today | YTD | 1 year | P/E |
|---|---|---|---|---|---|
| $221B | +1.77% | +25% | +37% | 24 | |
| $173B | +0.38% | +5% | +1% | 15 | |
| $143B | +0.61% | +16% | -1% | 38 | |
| $99B | +1.34% | +15% | +8% | 34 | |
| $89B | -0.41% | -28% | -30% | 25 | |
| $83B | +1.12% | +30% | +28% | 63 | |
| $76B | +1.16% | +20% | +6% | 45 | |
| $47B | +0.54% | -13% | -26% | 24 | |
| $19B | -0.48% | -41% | -43% | 24 |
Data center REITs$193.6B · +24% this year
Landlords for the buildings.
| Company | Market cap | Today | YTD | 1 year | P/E |
|---|---|---|---|---|---|
| $95B | +0.81% | +27% | +17% | 62 | |
| $65B | +0.60% | +14% | +1% | 84 | |
| $33B | +1.04% | +36% | +9% | 79 |
Software$935.1B · +2% this year
Companies selling AI to businesses.
| Company | Market cap | Today | YTD | 1 year | P/E |
|---|---|---|---|---|---|
| $468B | -1.11% | +11% | +9% | 172 | |
| $204B | -0.33% | -6% | -2% | 23 | |
| $162B | -1.39% | +4% | -15% | 98 | |
| $102B | -1.04% | -25% | -26% | 15 |
About the data
Capex is “purchases of property and equipment” from each company's cash flow statement, as filed with the SEC. It does not include equipment acquired through finance leases, which Microsoft and Oracle use for part of their data center capacity, so their totals are lower than the “capex including finance leases” figures they cite on earnings calls.
Quarterly cash flow figures are filed year to date, so each quarter is the difference between consecutive filings. Fiscal quarters are matched to the nearest calendar quarter: Oracle's fiscal year ends in May, Microsoft's in June and Nvidia's in late January.
The AI stack lists S&P 500 companies whose business is closely tied to data center spending. Inclusion is editorial, not a recommendation. Returns are weighted by market value. See the methodology for how stock data is sourced.
Frequently asked questions
How much are big tech companies spending on AI?
Microsoft, Amazon, Alphabet, Meta and Oracle spent a combined $566B on capital expenditures in the four quarters through Q2 2026, up 81% from the year before. Amazon spent the most, at $173B. Most of this goes to data centers, servers and chips for AI and cloud computing.
What is hyperscaler capex?
Hyperscalers are the companies that run the largest cloud and data center networks. Their capital expenditures, or capex, are what they spend on long-lived assets such as servers, networking gear, buildings and land. The spending is recorded as an asset and depreciated over several years rather than expensed right away.
Can hyperscalers afford their AI spending?
Over the past four quarters the five companies spent 82% of their operating cash flow on capex, leaving $126B in free cash flow. A rising share means less cash is left for buybacks, dividends and debt repayment, and some companies have turned to bond markets to help fund construction.
Which stocks benefit from AI data center spending?
Spending flows to chip designers such as Nvidia, Broadcom and AMD; memory makers such as Micron; chip equipment makers; networking and optical suppliers; server makers; and power and cooling companies. Together, the 45 S&P 500 companies tracked here make up 36% of the index by market value.