Big tech stunned the world last week when Amazon ($AMZN) and Google ($GOOG) both filed 2025 earnings reports and also announced plans to spend astonishing amounts of money on data centers in 2026. 

Their big bets came shortly after Meta ($META) and Microsoft ($MSFT) filed their own quarterly reports at the end of January, which also included plans for somewhat smaller but still staggering amounts of money going to data centers this year. 

The only one not yet disclosing fresh numbers is Oracle ($ORCL), but they’re scheduled to file their next earnings release on March 9, and we’ve already written about Oracle’s data center ambitions — and obligations — in the recent past. 

So what does the biggest picture look like? Do the hyperscalers even have the cash to cover all these capex costs? We cracked open our Multi-Company page to take a look, tracking capex and operating cash flow by calendar quarter and then adding up those numbers by year, even though Microsoft and Oracle use non-Jan. 1 fiscal years. 

Figure 1, below, shows the combined operating cash flow versus net capex spending for all five companies mentioned above, 2020 through 2025; plus estimated capex for 2026 (as per the companies’ guidance for 2026 capex spending). 

Notice that the spread between capex (in blue) and operating cash flow (in red) has been getting progressively narrower year after year. What we don’t know is estimated operating cash flow for 2026. (The hyperscalers have generally offered guidance on operating income, but that’s not the same.)

We could try to model an estimated operating cash flow by looking at the rate of increase from 2020 numbers ($262.5 billion) through 2025 ($602.83 billion). That rate changed from year to year: up 12.1 percent in 2021, down 0.3 percent in 2022, then up 34.6 percent the following year. 

The average rate of change in operating cash flow across that whole six-year period was 18.7 percent. If we assume 2026 operating cash flow is 18.7 percent higher than 2025 numbers, that implies a value of $715.56 billion. Which would imply a Figure 2, below, that looks like this:

That estimated 2026 differential is a lot narrower. Will it come to pass? We’ll have to wait and see. 

Individual Hyperscalers

Different individual companies tell different stories. For example, here’s the chart for Amazon ($AMZN):

Capex got dangerously close to exceeding operating cash flow in 2025. Then again, capex did exceed operating cash flow in 2021 and 2022, and Amazon is still here (although its share price did go through a marked decline in 2022).

On the other hand, here’s the same chart for Oracle ($ORCL):

A very different story. Oracle had solid operating cash flow over capex until 2024, and then capex soared, and then it soared even more in 2025, and it will soar even further in 2026. Plus, Oracle is a very different business than Amazon, which has always had large capex demands for its e-commerce operations. This is Oracle’s first venture into being a capital-intensive business.

Next we have Google ($GOOG), at a much more orderly progression:

Ditto for Microsoft ($MSFT):

And finally Meta ($META), or Facebook for the old-school purists:

Interesting that Meta also saw compression between operating cash flow and capex spending in 2025. It saw similar compression in 2022 — which, like Amazon, also coincided with a drop in share price over the year. 

The next question for financial analysts is how the hyperscalers will afford all this capex spending in 2026. They could squeeze cash flow even further, but they could also tap the debt markets. That’s what Oracle and Google have both done recently.

Can analysts track the debt that the hyperscalers are shouldering and then model the pressure those interest payments will add to net income, to better understand whether these AI bets are likely to pay off?

Yes, Calcbench lets you do that too. That will be in a future post.