Originally published on CNBC on 07-20-2026
Guest: Mohamed El-Erian, chief economic adviser at Allianz and president of Queens’ College, Cambridge
Intro
El-Erian argues the inflation fight is essentially won, explains why messy geopolitical news and calm markets are an unstable combination rather than a reassuring one, and puts a date on the AI overbuild he thinks is coming anyway.
Prediction Markets and Who Actually Pays for the Lottery Ticket
On the surge of retail participation in prediction markets, his read is mechanical rather than moral: “if you lower the barriers to entry to markets where there is significant upside, you will get a major reaction”
He agrees most participants are losing, and explains the gap with an airport comparison — the plane going to Las Vegas against the plane coming back, one carrying the upside and the other carrying the odds
What lowered barriers actually unlock is the illusion of edge: being able to bet on many different aspects of a game lets an individual think “I’m specialized in this area, I am going to prevail in this area”
On whether it’s good or bad long term, he splits the answer. The social implications are “not great,” and he points to the same pattern in lottery data: “Those who tend to bet are those who are least able to underwrite the losses”
But he does not land on restriction: “in general, I think access to markets are a good thing. And if you provide more access to markets, that’s a good thing”
Messy News, Stable Markets, Unstable Equilibrium
With fighting in Iran continuing and an expansion of attacks on both sides over the weekend, Brent was at 88 and crude at 82 on air, down from the low 90s the night before, with equities higher
The market’s working assumption, in his description, is that these escalations will be contained — and he treats the weekend as a live test of that assumption, which it passed
“And that combination is very striking. You normally don’t get that combination at all.” — on really messy news sitting alongside relatively stable markets
The stability is not the reassuring kind: “So this is a fascinating time because everything is in equilibrium. But it’s very unstable.” He names the 10-year Treasury and the yen as where he sees it
The Worst of the Inflation Is Behind Us
“I’m not into the we need three rate hikes. I don’t think we’re going to get any rate hikes. I think the worst of the inflation is behind us.”
His component-by-component case: tariff inflation is behind us, and most of the oil inflation is behind us
The one qualification he keeps live is energy prices, and he names the two he checks daily — regular gasoline above four dollars and diesel around five
On AI-driven inflation, he makes an explicit exception rather than an exemption: “The AI related inflation is inflation that I can live with because I truly believe there’s a productivity gain coming on that”
The Overbuild Is Coming, Just Not Yet
“there’s likely to be an overbuild because every innovation tends to overdo it” — he offers fiber as the reference case and says he could walk through others
Asked when he’d wake up and recognize it, he gives a number: “Probably in 3 to 4 years” — with the immediate caveat that it can run for quite a while before then
Where he thinks this differs from the fiber build-out is the absence of an end state. The tech people he speaks to believe it’s almost impossible to define where it ends, and in his framing “we simply don’t have the imagination”
His reason it has no natural end point is structural, not sentiment: “It is what James Manyika at Google calls the inventor of inventions” — recursive self-improvement that continuously allows more things to happen, which is what makes it very hard to predict
The contrast with laying track is deliberate: fiber had a defined build and a stopping point, and this is a build that keeps extending and keeps needing upgrading
El-Erian’s position is that the inflation risk that dominated the last cycle is largely spent and the AI capex cycle is the one worth watching instead — not because it’s fake, but because every innovation overdoes it in the initial phases, and this one has no visible end state to overdo it against.