ROI-AI frenzy stokes inflation heat too: Mike Dolan

(The opinions expressed here are those of the author, a columnist for Reuters.)

By Mike Dolan

LONDON, June 2 (Reuters) - The Iran energy story may be masking a bigger inflation worry. The AI boom is building a head of steam in prices under the surface, and it's a boom that will likely outlast any hiatus in the Gulf.

Everyone, not least the major central banks, is watching every twitch of crude prices around the Iran conflict for the cost-of-living hit. Yet the eye-watering AI investment frenzy is also stoking construction, factory activity and potentially prices, as bottlenecks and shortages in memory chips, tech equipment and software costs emerge worldwide.

This whopping business investment outlay is expected to exceed $800 billion this year alone and run into the trillions over the years ahead. It is already moving company earnings, projections and stock prices, as well as aggregate GDP. The direct effects on consumer prices are being pored over closely too.

Public concern about AI centers mostly on its long-term impact on jobs and the possible deflationary effects that could follow. Few have yet focused on the inflation pressure that likely comes first.

A curious and detailed Federal Reserve paper late last month, co-authored by then-Fed board governor and arch dove Stephen Miran, spotlights one area where the AI boom may already be fuelling consumer prices, before dissecting the readings to expose what may prove to be measurement errors.

It examined a sharp rise in the "Computer Software and Accessories" component of the Fed's preferred personal consumption expenditures (PCE) inflation gauge and threw up remarkable readings as the AI investment surge unfolds.

Although the category carries a relatively small weighting in the overall consumer price basket, it recorded an unprecedented rise between November and March.

The core goods component of the PCE rose 5.5% annualized over those four months, with software accounting for 2.8 percentage points of that increase, or more than half. The contribution was more than nine standard deviations above historical averages, it says.

The research made clear a large driver of this was a 70% increase in flash drives, or memory sticks, over a period when memory chip shortages were building in South Korea due to gigantic chip demand pressures around the global data center build.

The paper then argues that the prices recorded by statisticians may be misleading, unable to keep pace with rapid technological development and quality upgrades, and possibly imputed incorrectly.

The PCE calculation draws inputs directly from the consumer price index (CPI), and a mismatch between PCE and CPI categories alone may account for a quarter of the outsize gain. Together with other measurement errors, up to half the gain may be overstated.

HOT AND COLD

Not everyone is ready to dismiss these pressures as statistical noise.

Whatever the doubts about methodology, the price rises are very real. And that is before considering how AI-driven input costs ripple through other chip-intensive products, from autos to energy.

"The AI Capex Boom is inflationary," Deutsche Bank strategist George Saravelos said in a note on Monday. "There is no compelling evidence of an impact on the labor market but a lot of evidence of demand-side inflation. AI is currently putting upward, not downward, pressure on rates."

Morgan Stanley's baseline differs and its view of U.S. inflation is relatively sanguine compared with an increasingly hawkish market take.

Even so, the bank's alternate scenarios flag the risk that AI heat feeds price rises more aggressively, including the possibility that a "combination of AI-related price pressures through categories like chips, memory, semiconductors, and software means core inflation firms and the Fed hikes."

What's more, Morgan Stanley also outlines the possibility that AI "animal spirits" tighten the labor market as the broader economy accelerates.

In that scenario, headline inflation could end the year where it is now at just under 4%, with core inflation picking up - likely prompting the Fed to consider up to 100 basis points in hikes. Morgan Stanley puts the probability at 15%.

OK, so there's a lot of "ifs" and "buts" in these alternate scenarios.

But the consensus carries its own questionable assumptions.

What's more, the AI debate dwells increasingly on the true cost of AI computing itself, with a growing focus on tracking AI tokens that measure that cost and possible re-pricing of related contracts down the road.

If the cost of producing and using AI is higher than currently assumed, that could play out several ways: costs either passed through products, recouped via job shedding, or simply stifling the use of AI overall.

For the public and policymakers, the risk is that the whole scenario - not unlike the Iran-related energy shock - delivers a one-two punch: inflation spikes for a protracted period, only to reverse at some point over the horizon as demand destruction sets in with job losses.

Societe Generale's global chief economist Wei Yao called it "the convexity of central bank tightening" in applying the sequence to Iran and oil, but it could extend to AI fallout too.

Heat followed by cold?

(The opinions expressed here are those of Mike Dolan, a columnist for Reuters.) Enjoying this column? Check out Reuters Open Interest (ROI), your essential new source for global financial commentary. Follow ROI on LinkedIn, and X. And listen to the Morning Bid daily podcast on Apple, Spotify, or the Reuters app. Subscribe to hear Reuters journalists discuss the biggest news in markets and finance seven days a week.

(by Mike Dolan; Editing by Marguerita Choy)