Mark Zandi, chief economist at Moody’s Analytics, went on CNBC this morning with a take that should land uncomfortably for anyone betting on rate cuts before year-end. The headline number is bad enough. May PCE inflation came in at 4.07% year over year, the hottest reading since April 2023, with core PCE at 3.41%. Zandi thinks the top-line number has probably peaked. The reason he thinks that should worry you more than the report itself.
Why energy is driving the headline lower, and why that relief is temporary
Zandi’s read on the headline number was straightforward. “I think it’s the peak. Oil prices are way in and gasoline prices are coming in… that 4% plus is probably the peak,” he said. You can see why he says it. Energy prices jumped 24.26% year over year in May, with a 4.03% monthly move on top of an 11.58% spike in March. That kind of volatility burns off. Gasoline rolls over, the headline rolls with it.
But Zandi was careful to separate the cyclical relief from the structural problem underneath. “I do think it’s a long road back to the Fed’s target though. We’re double that. And there’s a lot of other things going on,” he said. Double the target is the operative phrase. Headline PCE is 2.07 percentage points above the Fed’s 2% goal, and core sits 1.41 points above it. The path back is not a quarter or two of soft data.
The AI inflation channel nobody priced in
This is where Zandi got specific, and where the framing differs from the consensus narrative that AI is a deflationary force. “Broadly artificial intelligence, AI is juicing up inflation, and you can see it in the higher cost for chips… prices for almost all consumer products are going to go up,” he told CNBC.
Apple (NASDAQ:AAPL | AAPL Price Prediction) and other hardware makers are absorbing chip cost increases driven by AI demand, then passing them through. The pass-through shows up in the goods data. Goods inflation accelerated to 4.78% year over year in May, up from 1.8% in February. Goods inflation more than doubling in three months is not a chart most economists had on their bingo card for 2026.
Then there’s the electricity bill. “Electricity prices are also up. And that traces back… to AI… that’s going to be a pressure on inflation,” Zandi said. Hyperscaler data centers pull power on a scale the grid was not built to serve cheaply.
Households compete with NVIDIA (NASDAQ:NVDA)-stuffed buildings for the same electrons, and the auction price goes up. Services inflation reflects some of this stickiness already. Services prices rose 3.76% year over year in May and have held in a 3.36% to 3.76% range for six straight months.
The productivity payoff has not arrived yet
Bulls on AI generally argue that the technology will eventually be net deflationary, with productivity gains crushing input costs. Zandi agrees with the destination. He disputes the timeline. “One of those productivity gains actually kick in from AI. They haven’t so far… underlying productivity growth is 2%-ish… the average… since World War II,” he said. AI productivity gains are expected to materialize over the next 5 to 10 years, on his read. Until then, you get the cost side without the offset.
The sectoral data backs the wait-and-see view. Information sector growth decelerated to 1.5% in Q1 2026 from 3.2% in Q3 2025, and manufacturing slowed to 1.3% from 3.2% over the same window. If AI were already lifting output per hour broadly, those lines would not be sliding. For the official series and methodology, the BEA’s PCE price index page is where the monthly release lives.
So the working framework Zandi gave you. Headline PCE probably tops here. Core grinds higher. The Fed’s 2% target stays out of reach while chips and kilowatts both inflate. And the AI offset that ends this story arrives sometime in the back half of the decade, give or take.