6 Hyperscalers Driving the AI Revolution Are Projected to Spend $1.3 Trillion on Capex in 2027. Only 1 Is Forecast to Have Positive Free Cash Flow.

Hyperscalers driving the artificial intelligence (AI) revolution have been spending hundreds of billions annually since 2024 to build AI infrastructure.

This includes data centers equipped with various chips, memory, servers, and more to continue powering the insatiable demand for AI, whether from consumers using large language models (LLMs) or companies building out AI solutions or integrating AI into their existing businesses.

Next year, that number is expected to rise to $1.3 trillion from just six of the major hyperscalers, according to a new report from S&P Global, which provides grades on the debt issued by most major companies. Only one of these companies -- Microsoft (NASDAQ: MSFT) -- is projected to have positive free cash flow (FCF) next year.

Image source: Getty Images.

AI spending growth is still expected to be robust

Besides Microsoft, the five other companies included in S&P's report were Alphabet (NASDAQ: GOOG)(NASDAQ: GOOGL), Amazon (NASDAQ: AMZN), Meta Platforms (NASDAQ: META), Oracle (NYSE: ORCL), and Space Exploration Technologies (NASDAQ: SPCX) (SpaceX for short).

Collectively, these companies had $470 billion in capital expenditures (capex) in 2025, along with a forecast for $870 billion this year and a projected $1.3 trillion next year. If S&P Global is correct, that means these six companies will increase AI capex by about 50% next year.

That's not as high as this year's increase, but it's still impressive, given how much larger the numbers will be. On recent earnings calls, most of the CEOs of these companies said they expect strong capex growth in 2027.

However, this year, the incredible spending has begun to deteriorate the balance sheets of these tech titans, which is hard to fathom, particularly for Alphabet, Amazon, Meta, and Microsoft, which have for years generated phenomenal free cash flow (FCF) and earnings.

All the CEOs of these companies have defended this spending, claiming that it will generate compelling returns and that not committing this capital would be an even greater risk by failing to keep up with a technology that could very well change society as we know it.

But institutional investors are skeptical and, at times, have not been buying these stocks despite some very strong AI-related revenue growth. In the second quarter of this year, three of these companies still generated free operating cash flow. Next year, if S&P Global is correct, only one will.

Here is each company's projected 2027 capex and FCF:

Source: S&P Global.

It's worth pointing out how much more S&P Global projects Alphabet and Amazon to spend than Microsoft in 2027. This is particularly interesting because these companies control the three largest cloud businesses in the world, which will play an enormous role in powering AI computing.

S&P Global says that Microsoft relies more on leases than other hyperscalers do, which could allow the company to reclassify certain finance leases included in capex as operating leases that flow through the income statement and thus improve FCF.

Microsoft has a complex accounting situation

S&P Global says that Microsoft has over $329 billion in future lease obligations, ahead of Meta, the next largest at $279 billion. This could explain why S&P projects Meta, the major provider of computing capacity, to have significantly lower capex next year than Alphabet and Amazon.

As hyperscalers continue to build more data centers, accounting has become a point of contention, especially as they set up off-balance-sheet arrangements. It's hard to know how the market will perceive these accounting moves, or if others will look to follow Microsoft, as high capex is a major part of the bearish argument for the hyperscalers right now.

Still, S&P Global continues to assign Microsoft debt the highest possible credit rating, AAA, making it one of two publicly traded companies -- along with Johnson & Johnson -- with such an elite rating.

S&P Global also says that its models generally assume an inflection in 2028 for the hyperscaler group, with capex flattening out and revenue accelerating, which could lead to a return to positive FCF. But this is one of the big questions investors are trying to figure out: whether this actually happens and leads to strong FCF and strong returns on this huge amount of invested capital.

Bram Berkowitz has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Alphabet, Amazon, Meta Platforms, Microsoft, Oracle, and S&P Global. The Motley Fool recommends Johnson & Johnson. The Motley Fool has a disclosure policy.