Three Fed Officials Just Voted to Raise Rates. The Real Fight Is Whether AI Is Inflationary.

The Fed did nothing today, which was expected, and the way it did nothing was the story. Rates stay at 3.50% to 3.75%, a fifth straight meeting without a move.^[1] The vote was 9-3. All three dissenters, Beth Hammack of Cleveland, Neel Kashkari of Minneapolis and Lorie Logan of Dallas, wanted to raise the target range by a quarter point today.^[1]

That is the most dissents at a single FOMC meeting since September 2016.^[2] And every one of them points the same direction, which is the part worth sitting with. This is not a committee split between doves and hawks. It is a committee where the only people willing to break consensus are the ones who think policy is too loose.

Why Stocks Fell on a Decision Nobody Was Surprised By

The Dow closed down about 1.5%, the S&P 500 and Nasdaq each about 0.6%.^[2] If the hold was expected, why sell?

Because the rate was never the news. The composition of the vote was. A unanimous hold says the committee is comfortable. A 9-3 hold with three members voting to tighten says the committee is one or two inflation prints away from moving, and it tells you which way. Markets did not reprice today's rate. They repriced September.

This is the thing about Fed watching that trips people up. The decision is almost always priced in by the time it arrives. What moves markets is the distribution of opinion behind it, and today that distribution shifted hawkish in a way the futures market had only partly accounted for.

What the Statement Actually Said

Two sentences carry the weight. On prices: inflation “remains elevated relative to the Committee's 2 percent goal,” with part of it attributed to supply shocks in energy and other sectors.^[1]On jobs: “Job gains have kept pace with the workforce, and the unemployment rate has changed little.”^[1]

Read those together and the dissenters' case writes itself. The Fed's mandate is inflation and employment. Employment is fine. Inflation is above target. If one side of the mandate is satisfied and the other is not, the argument for holding has to rest on something other than the data in front of you, and what it actually rests on is a forecast that the pressure is temporary.

The word doing the work is “supply.” A central bank can do very little about supply shocks. Raising rates does not produce more energy or more transformers. So attributing inflation to supply is, in effect, an argument for patience: this is not demand you need to destroy, it is a bottleneck that will clear. The dissenters are betting that some of it is not a bottleneck.

The Question Underneath: Is AI Inflationary?

Here is the detail that got almost no coverage and is the most interesting thing to come out of this meeting. Warsh has stood up a set of task forces, reporting from late 2026 onward, to examine whether artificial intelligence is raising growth without raising prices, and whether the Fed is thinking about inflation correctly at all.^[2]

Central banks do not casually announce that they are unsure how to classify the largest capital investment cycle in the economy. That is a real admission, and it is the correct one, because AI genuinely pushes in both directions and the two effects arrive on completely different schedules.

The inflationary channel is happening already.Building data centers bids up electricity, gas turbines, transformers, high bandwidth memory, construction labor and electricians, in specific regions, faster than supply can respond. That is demand meeting an inelastic supply curve, which is the textbook definition of a price increase. And it is concentrated exactly in the “energy and other sectors” the statement gestures at.

The disinflationary channel is a forecast. If AI raises productivity broadly, the same output costs less labor, and that pushes prices down across the economy. That is the optimistic case and it may well be right. It is also, so far, mostly not visible in the aggregate data, which is the whole reason the task forces exist.

So the Fed is holding rates in the gap between a cost it can measure and a benefit it can only project. That is a defensible thing to do. It is also precisely what the three dissenters object to, because if the productivity gains arrive late or small, the Fed spent 2026 accommodating an inflation impulse it had described as temporary.

How the July 4 Call Aged

Three weeks ago I wrote that markets had heard Warsh say inflation risks had eased and had translated that into a coming rate cut, and that those are not the same thing. That has aged well, and not because of any great foresight. It aged well because the market was doing something it does structurally: it collapsed a probability distribution into the single outcome it wanted.

The distance between “risks have eased” and “we will cut” turned out to be large enough that the live debate today was about a hike. Not a cut arriving later than hoped. The opposite direction entirely.

Warsh's deliberate retreat from forward guidance is what makes this keep happening. Previous chairs told markets where policy was going; Warsh has made a point of not doing that. The predictable result is that markets fill the silence with their own preference, and then get repriced when the committee does something else. If you are positioning on Fed expectations, that gap is the risk, and it is now a structural feature of this Fed rather than a one-off.

What Actually Changes From Here

September is the live meeting. The committee will have the July and August CPI reports by then.^[2] The June projections already penciled in one quarter-point increase by the end of 2026.^[2] So the hold is not a decision that rates stay here. It is a decision to wait for two more data points before delivering something the committee has already sketched.

Watch the split, not the level. If September comes with three dissents again, or four, the hike is arriving whatever the statement says. If the dissents fall to one, the supply-shock reading won and patience is the policy.

And watch what the bond market thinks of all this. The curve is currently pricing modest future cuts while the committee argues about a hike, which is a disagreement rather than a signal. The yield curve just missed a recession call for the first time, so it is worth knowing how much weight that pricing deserves.

Mortgage rates will not track this cleanly.Worth repeating every time the Fed moves or doesn't, because it is the single most common misreading in personal finance: mortgage rates follow the long end and the spread, not the federal funds rate. A hawkish hold can coincide with mortgage rates falling.

For the valuation side of the same question, the bull and bear cases are both laid out in is this an AI bubble, and the specific bottlenecks that make the build-out inflationary are in the shift from picks-and-shovels to power.