This hidden Microsoft metric bodes well for Nvidia

There are plenty of reasons why investors are excited Nvidia (NVDA 3.37%). The chipmaking giant has been the big winner in building artificial intelligence (AI) infrastructure, which has helped it carve out a big moat and become the dominant supplier of graphics processing units (GPUs) used in AI servers.

One thing that has kept the stock from performing even better is that some investors worry that after an initial surge, spending on AI infrastructure could slow from current levels. However, there is hidden value found within Microsofthis (MSFT -0.14%) recent annual 10-K filing, which should allay any worries that even brighter days are ahead for Nvidia.

Image source: Getty Images.

Financial leasing for data centers

Buried in the footnotes of its quarterly and annual reports, Microsoft lists the number of finance leases, mostly for data centers, that have not yet begun. The details include the amount of money he contracted for leases that he has yet to start using.

At the end of June, these finance leases that had not yet started totaled $108.4 billion. To put this into context, Microsoft had $27.1 billion in total finance lease liabilities at the end of fiscal 2024 (which ended in June), and finance leases that had not yet commenced at the end of the fiscal year 2023 were 34.4 billion dollars. So the value of the finance leases it has taken out that haven’t started yet has tripled in the last year.

These leases are expected to commence between fiscal years 2025 and 2030. They will have terms ranging from one year to 20 years.

So what does this all mean? Well, finance leases are usually long-term agreements where the owner of an asset gives control of it to another party in exchange for payments. Typically, at the end of the lease, the lessee (the party making the payments) has the option to purchase the asset for a nominal amount.

Microsoft has made it clear that these leases are for data centers, meaning it has contracts in place to spend a lot of money on data centers in the coming years. Now, some of that could be through partnerships with Oracle and CoreWeave, as speculated earlier UBS. But that still leaves plenty of new space in the data center set to be added.

How will this affect Nvidia?

For Nvidia, this is just another indication of future data center spending — and by extension, GPU spending — going forward. Hyperscalers like Microsoft are already spending an enormous amount of money on capital expenditures (capex), much of which is directed toward AI and data centers. In fiscal 2024, Microsoft spent $44.5 billion in investments, almost entirely on cloud and AI, and said it will spend even more in fiscal 2025.

Half of the investment last fiscal year was on infrastructure to build and lease data centers, and the other half was on server CPUs and GPUs. These not-yet-started finance lease obligations point to even more expenses in the future and the need for more GPU servers.

And Microsoft isn’t the only company spending a lot of money on AI-related investments. Amazon spent $30.5 billion in capital expenditures in the first six months of the year and said it would spend more in the second half, most of it to support the growing need for AWS infrastructure.

Alphabetmeanwhile, it spends more than $12 billion per quarter on capital expenditures, with the largest component for servers and Meta platforms said it expects a significant increase in investment growth in 2025 from the recently revised upward budget for 2024, from $37 billion to $40 billion.

These big tech companies are racing to build their AI infrastructure, and there seems to be no end in sight. Large language models (LLMs) require more and more computing power to train as they advance, which means not only more GPUs, but exponentially more. For example, Elon Musk-backed xAI trained its next Grok-3 LLM using 100,000 GPUs, five times more than the Grok-2, while Alphabet said its Llama 4 model would have require up to 10x more GPUs than the 16,000 used for Blade 3.

All of these data points continue to point to a lot of spending ahead on AI infrastructure and GPUs, all to Nvidia’s benefit.

Is it too late to buy Nvidia?

While Nvidia stock has been a huge winner, the stock is attractively valued given the huge opportunity that still appears to be in front of it. It trades at a forward price-to-earnings (P/E) ratio of only about 29, based on analyst estimates next year, and a price-to-earnings-growth (PEG) ratio of under 0.8. A PEG below 1 is usually considered undervalued, and growth stocks will often command PEGs well above 1.

NVDA PE Ratio data (Before 1y) by YCharts

As such, I think the best approach to take regarding Nvidia is to listen and watch what Nvidia’s biggest customers are saying and doing. On that front, it looks like those customers plan to spend a lot more money on GPUs from the company in the coming years, with Microsoft aiming to be one of the biggest spenders. With Nvidia stock still attractively priced, I’d be a buyer.