Warsh's Opening Move Was Hawkish. The Bond Market Isn't Fully Convinced.

On June 17th, the two-year Treasury yield jumped 16 basis points in a single session. That was the largest single-day move on an FOMC meeting day since March 2008. March 2008 — the month Bear Stearns collapsed and the Fed slashed rates 75 basis points into a crisis it had spent months pretending wasn't coming. The comparison is imprecise but the signal is not: a 16-basis-point move on a day the Fed did nothing is the bond market telling you it was caught off guard. And bond markets are rarely surprised by things that don't matter.

What surprised them was Kevin Warsh.

New chairs get scrutinized for tone more than content. The content of the June statement was predictable enough — hold at 3.50%–3.75%, acknowledge inflation, project one hike before year-end. What the market didn't expect was the temperature. Warsh's first press conference carried the quiet firmness of someone who has already decided and is now informing you, rather than deliberating in public. Nine of eighteen FOMC members put a dot above the current range for year-end. Warsh himself declined to submit a dot, saying it wouldn't be "helpful in the conduct of policy" — which could mean he wants to preserve optionality, or could mean he doesn't want a number that constrains him when the data forces his hand in September. Either way, the absence of a dot from the chair is itself a signal. A different kind of dot. A floating one.

The PCE data, released Thursday, landed broadly in line with where those dots were pointing. Headline at 4.1% year-over-year, the fourth consecutive monthly acceleration, the highest reading since April 2023. Core at 3.4%, a tick above consensus. Services prices up 0.5% month-on-month — restaurants, hotels, healthcare, auto repairs. The stuff that doesn't come down when Hormuz reopens. Oil falling 3% on Friday to around $69 a barrel will eventually bring headline PCE lower, and the consensus view is that May was the inflation peak. That's probably right. But the consensus view also said April was the peak. And March before that.

This is the trap Warsh inherited and the trap he is now visibly trying to navigate without triggering.

The structure of this inflation is not the structure of 2022. That was demand-pull amplified by supply chain rupture, a unique collision of fiscal excess and pandemic logistics. This is messier — an energy shock from a war in the Middle East seeping into services through secondary price-setting, while AI infrastructure buildout creates genuine goods-side supply constraints (Apple raising iPad prices due to memory shortages is not something the Phillips Curve has strong opinions about). The Dallas Fed's trimmed mean PCE, which strips the tails, sits at 2.4% over the past twelve months. The underlying economy, structurally speaking, is not running especially hot. What's running hot is a specific, war-driven energy impulse that is now, gradually, reversing — and a goods shortage in semiconductors that is itself a consequence of the most ambitious infrastructure investment cycle in a generation.

Warsh has said, explicitly, that supply-shock inflation should generally be looked through. He has also said AI will be disinflationary through productivity gains. Both of these things may be true. The problem is that you cannot say them out loud right now. The moment the Fed chair signals that he views the current inflation as transitory by another name, financial conditions ease, the dollar falls, risk assets rip, long yields decline, mortgages get cheaper, spending accelerates, and the inflation that was supposed to be looking-through-able becomes embedded in expectations. The credibility trade requires him to act more hawkish than he may privately believe is warranted. That's the debt he's paying to inherit an institution whose credibility was questioned under the previous regime.

Markets are pricing three hikes this year. Deutsche Bank wants two, with the federal funds rate ending 2026 at 4.1%. The probability of a September move is running around 62–63%. The 10-year Treasury is trading at 4.39%, down about 7 basis points on the week as oil fell and the PCE didn't come in hotter than feared. The 30-year is at 4.86%. The curve has steepened from its inverted lows — 10-year versus 2-year spread is now +27 basis points — which means the bond market is repricing not just near-term policy but the long-run neutral rate. A steeper curve in an environment of rising inflation and firm growth is not a recession signal. It's the market telling you that money is going to be more expensive for longer than the Fed's own prior projections implied.

That repricing has costs that are still working through the system slowly. The Bloomberg US Aggregate Bond Index has returned 0.38% year-to-date. Coupon income is roughly covering price losses. The AI hyperscalers — Microsoft, Amazon, Alphabet, Meta — have issued hundreds of billions in new debt to fund data center construction, at investment-grade spreads that were already slim to begin with. Fidelity's bond team has flagged this explicitly: the capex intensity of the AI buildout could stress credit ratings of the very issuers whose bonds have been treated as quasi-risk-free by the market. A bond can't compound the way a stock can if the AI bet pays off. But it can lose value if the issuer gets downgraded. The credit market has not repriced this risk. Not yet.

The historical analogy the bears reach for is 1979, and it's not wholly inappropriate. Arthur Burns spent the 1970s stopping and starting, allowing inflation to become a cultural expectation, before Paul Volcker arrived and inflicted deliberate economic pain sufficient to break it. The lesson extracted from that episode — and the one that anchors every serious monetary economist of the past four decades — is that credibility, once lost, is extraordinarily expensive to rebuild. Warsh knows this. His June performance was a down payment on that credibility. The 16-basis-point move in the two-year was the market revaluing its assessment of him upward.

But a down payment is not a mortgage. The real test comes in September when the PCE data may well be running lower, oil may be in the $60s, and the political pressure to hold — from the White House, from the housing market, from every rate-sensitive constituency that exists — will be substantial. Cutting or holding in September, if inflation is visibly decelerating, is defensible on the data. The question is whether Warsh will cut or hold for the right reasons — because the supply shock is genuinely dissipating — or whether he will quietly declare victory on inflation while the services component remains sticky and the trimmed mean slowly drifts back up. The first is credible central banking. The second is Burns with better tailoring.

The bond market's 62% September hike probability says it doesn't yet know which one it's dealing with.

Neither do we.