When the ‘AI Inflation’ Hits the Fan
Rising costs associated with AI are weighing on stocks like Apple. But there’s a bigger question looming for hyperscalers.
You've reached your free article limit
You've read 0 of 1 free Pro articles.
Micron Technology (MU) posted blowout quarterly results after the bell on Wednesday. That was good for an over 15% rally in the stock on Thursday. Unfortunately, Micron’s results did not take the rest of the Nasdaq with it in trading yesterday. Micron signed 16 strategic customer agreements during the quarter. These three-to-five-year agreements are for customers across the data center, consumer, and auto market segments. They now account for roughly 25% of overall revenue. Stifel Nicolaus noted that the terms of these new long-term supply agreements are exceptional. They also signal a major paradigm shift with clear positive implications for Micron.
While these agreements might help get Micron away from the boom-and-bust cycles it always has had to navigate through, they are not necessarily good for their primary customers. They now have to deal with higher costs which is all part of “AI Inflation,” a buzzword that seems to be gaining traction recently.
Economist Mark Zandi noted this week that AI is “juicing” up inflation instead of ushering in deflation as some expected, because of rising component and electricity costs.
Rising AI inflation is one of several reasons that the Magnificent Seven have underperformed recently. These rising costs are forcing Apple (AAPL) to hike prices for some of its core products like the iPhone significantly. Notably, the stock of Apple fell 6% on Micron’s results yesterday. The stock’s worse daily decline in over a year.
Higher memory prices are just one headwind the hyperscalers are facing right now as they face higher costs for components. The huge increases to their capital budgets are having substantial impacts on myriad fronts. Stock buybacks have slowed to a trickle to what they were a year or two ago. Debt issuance has exploded, with Bank of America boosting its projection substantially in mid-March to $175 billion in debt issuance across the major hyperscalers in 2026. This is probably not the last time this estimate will be revised up this year.
And this does not account for all the off-balance sheet obligations these entities are incurring. These include special purpose vehicles, or SPVs, being used to finance these huge data center buildouts. Nor do they include future lease obligations for these facilities that have been agreed to but have not started yet. These amounted to $183 billion at Meta Platforms (META) at the end of Q1.
And even with debt issuance and growth in off-balance sheet liabilities, the demand for funding for AI capital expenditures seems insatiable. Alphabet (GOOG) just raised an additional $85 billion this month via equity offerings and private placement. Space Exploration Technologies Corp. (SPCX), which just raised $85 billion of capital from its record IPO on June 12, has already come back to access the capital markets with a $25 billion bond sale.
In summary, there is no doubt the suppliers of the “picks & shovels” will continue to prosper mightily from the huge increases in AI capital expenditure budgets. But with AI inflation continuing to push AI infrastructure costs higher, whether the hyperscalers will end up ever generating an acceptable return on their massive investment is an increasingly open question.
At the time of publication, Jensen had no position in any security mentioned.
