How AI Spending Is Affecting Balance Sheets

Last month we had a post about capex spending among S&P 500 firms, and how capex spending seems to be rising briskly, but that spending is actually being driven by a few tech giants spending gobs of money on data centers for artificial intelligence. 

Today we want to revisit that issue from a different angle: how is all that spending changing the nature of the tech giants’ balance sheets? 

For many years, those balance sheets were notable for two basic traits: (a) lots of cash; and (b) lots of goodwill or other intangible assets. Physical assets — land, buildings, and equipment typically listed under the Property, Plant, and Equipment line item — accounted for a relatively small part of the tech giants’ overall assets. 

Well, that’s changing.

Figure 1, below, shows the ratio of “PP&E” assets versus total assets for four tech giants leading the AI arms race: Amazon ($AMZN), Google ($GOOG), Meta ($META), and Microsoft ($MSFT). 

As you can see, the ratio for all four firms has been increasing steadily since late 2023, roughly one year after the debut of ChatGPT. (ChatGPT is owned by OpenAI, but Microsoft provides the bulk of the computing infrastructure for it.) 

It’s particularly striking to see the swift increases for Microsoft (in green) and Google (in blue), since their AI systems are the two largest. In contrast, Amazon (in red) has seen less of an increase, but started from a higher base. Then again, Amazon’s PP&E includes all the warehouses, vehicles, and other equipment for its e-commerce operations, so its PP&E spending is unlike those of the other three tech giants. 

Figure 2, below, shows the ratio of goodwill to total assets for the same time period.

This chart tells a very different tale: that goodwill is declining as a portion of total assets. The exception is Microsoft and that gigantic spike in goodwill at the end of 2023, but that coincides with Microsoft’s acquisition of Activision-Blizzard, with boatloads of goodwill in tow.

And for a company-specific perspective, we offer Figure 3, below. This one shows how several important line-items on the balance sheet of Microsoft have evolved over time. 

Again we see that big jump in goodwill, as well as a big jump in cash, in latter 2023 as Microsoft digested the Activision-Blizzard deal. Overall, however, the big message in Figure 3 is the larger and larger red line, which represents PP&E. 

That line-item went from $74.4 billion in Q2 2022, to $135.6 billion in Q2 2024, to $205 billion in Q2 2025. And while Figure doesn’t include all assets, among the four assets it does track, PP&E went from 43.5 percent of the total at the start of 2022 to 54.3 percent in Q2 2025.

In other words, the AI arms race is slowly but surely transforming the tech giants from primarily light assets (goodwill, intellectual property, and other intangible assets) to increasingly heavy assets (PP&E). That has important implications for free cash flow and other performance metrics, which in turn has important implications for financial analysis and investment decisions. 

Calcbench subscribers can find this data in several ways, all of them easy:

• The Company-in-Detail page lets you see one specific company’s primary financial statements, including the balance sheet, in line-item detail. (Here’s the Microsoft example.) You can then look back through prior periods as much as you’d like.

• The Multi-Company page lets you see specific disclosures from one or more companies; first looking at specific periods and then expanding to a time-series of data if that’s your cup of tea. (Again, the Microsoft example.)

• Our Bulk Data Query page lets you pull together aggregate data from potentially huge groups of companies. This is great if you want to see how Figure 3 might look in aggregate for, say, the entire S&P 500 or all companies in the tech sector as defined by SIC code.

• Or, for power users with their own advanced analytics capability, you can always use our API to export our data directly into your own models and data feeds. Email us at us@calcbench.com for details.

And the results for all of the above pages can always be exported quickly and easily to Excel!