How to Approach Uncertainty Ahead of Nvidia’s Earnings Release

According to checks conducted through multiple channels by an analyst at TD Cowen, Microsoft has canceled lease agreements for several AI data centers it had planned to build. The regions where cancellations have been confirmed are mainly places like Arizona in the US, where AI computing demand outstrips available supply.

The analyst pointed to the Jevons paradox—which we also discussed in the Aware Original piece “Google Cloud’s Earnings Miss: Why It Means Buying More GPUs”—and argued that this move is “at odds with the company’s stated intention to carry out its $80 billion capex plan for this year as scheduled”. The analyst maintains the view that “the surge in AI computing demand after the launch of DeepSeek’s new model in China represents temporary excess demand created by running DeepSeek’s model in parallel with existing models”.

In an official statement, Microsoft said, “We may strategically adjust our infrastructure in certain regions, but we continue to see strong growth across all regions,” adding that “our planned $80 billion in capex this year to meet strong customer demand is progressing smoothly.”

Jordan Klein, an analyst at Mizuho Securities, commented on the TD Cowen analyst’s remarks, saying, “To me, this just looks like part of normal day-to-day operations,” and added that “a massive company that spends $80 billion a year (roughly 113 trillion won) on capex can adjust data center lease agreements as needed, and most of the canceled leases were not even fully executed contracts.”

It is hard to say whether negative headlines are being highlighted because share prices are falling, or whether share prices are falling because of the negative headlines. What is clear is that the stocks that led last year’s rally—big tech and AI-related names—are now trending downward. Shares of Palantir (PLTR), known as an AI software stock, have dropped 24% over the past five trading days.

I see this back-and-forth debate and the recent pullback in share prices as a normal correction that occurs as market participants make a more sober assessment. The reality is that some companies labeled as AI plays or AI beneficiaries have little to do with AI or are unlikely to benefit from it, or have no real technological differentiation or competitive edge, yet still saw their share prices soar.

With Nvidia, the bellwether of AI infrastructure, set to report earnings on February 27 US time, this may also be a correction aimed at clearing away some of the heightened tension and uncertainty. It is almost taken for granted that Nvidia’s results will be strong, so investor attention is likely to focus on management’s commentary about the outlook for the market going forward.