Big Tech earnings season and the capex spiral

Big Tech earnings season starts next week with Google, but the most important number won't be profit or revenue. It will be how much these companies say they'll spend on AI data centers.

Google, Amazon, Microsoft, and Meta have already laid out plans to spend more than $700 billion this year. Investors know the companies are racing to add computing power. The harder question is how much extra capacity those dollars actually buy.

The answer is getting worse. Memory chip prices have surged, while power equipment, construction materials, skilled workers, and electricity connections are all harder to secure.

Morgan Stanley now estimates the cost of building one gigawatt of AI capacity has risen about 20% for several leading systems. One common Nvidia-based setup, for example, has gone from roughly $29 billion to $35 billion per gigawatt; a newer version has climbed from $41 billion to $49 billion.

That creates a nasty loop. Big Tech orders more AI data center gear. Shortages worsen. Prices rise. The companies then raise spending forecasts partly to cover those higher prices, which creates still more demand and even higher prices.

Brad Gastwirth, head of research at Circular Technology, estimates about 20% to 30% of the next increase in AI capex will reflect inflation, while 70% to 80% will still represent real expansion. In other words, higher spending doesn't always mean the AI buildout is accelerating at the same rate.

"Investors need to be careful when interpreting higher capex. There is absolutely an inflation component," Gastwirth told me recently.

That distinction matters to investors. Earlier research found that soaring memory prices could explain about 45% of the growth in capex by the big cloud companies this year.

Cantor Fitzgerald analysts expect little change to 2026 capex plans this earnings season, but see 2027 estimates moving sharply higher, including $283 billion for Google, $271 billion for Amazon, and $200 billion for Meta.

So, will anyone blink and stop increasing capex? Probably not yet. No company wants to look cautious while AI rivals are racing ahead. But listen closely to what accompanies any spending increase this time.

"When the companies discuss capex, I'd pay close attention to whether they're also talking about power capacity, GPU deployments, memory purchases, networking, and new data center campuses," Gastwirth said. "If capex rises alongside those metrics, it points to genuine expansion rather than simply higher costs."

Without those details, a larger capex number may simply mean Big Tech is paying more to stand still.

Sign up for BI's Tech Memo newsletter here. Reach out to me via email at [email protected].