The Warsh Dove Trade. An Autopsy.
INTERNAL — RATES & MACRO DESK
Re: The Warsh Dove Trade. An Autopsy.
Monday, June 22
Let's get this done quickly so we can move on.
The thesis was clean. Too clean, in retrospect. Trump nominates Kevin Warsh — a man who spent the better part of the past two years writing op-eds in the Wall Street Journal about artificial intelligence as a "significant disinflationary force," a man whose entire intellectual pitch to the Senate confirmation committee hinged on AI productivity unlocking a rate-cut cycle, a man that the sitting president selected explicitly because he expected him to bring rates down before the midterms. The trade writes itself. Long duration. Short the dollar. Cover your rate hike hedges. The new chair is a dove in hawk's clothing, and the clothing is coming off by Q3.
Wednesday happened.
In 40 minutes of press conference, Warsh erased the thesis entirely. The 2-year yield gained more than 16 basis points to 4.216% before he'd finished answering the second question. The Dow fell 507 points. The S&P closed down 1.21%. Ed Yardeni — who had built a reasonably coherent version of the dovish Warsh case — said he was "blown away." Scott Clemons at Brown Brothers Harriman called it "regime change, but in a velvet glove." The velvet part is generous.
Here is what actually happened at the June 17 meeting, stripped of narrative comfort.
Nine of eighteen FOMC officials now pencil in at least one rate hike before year-end. Six of those nine project two hikes. The median dot moved from 3.4% in March to 3.8%, a flip from implying a cut to implying a hike relative to the current 3.625% midpoint. PCE inflation was revised up to 3.6% for 2026 — this is the Fed's own forecast, not some outside bear case. The statement itself was shortened, simplified, and stripped of every phrase that markets had been reading as an easing bias for the better part of eighteen months. And Warsh told reporters, plainly, that forward guidance is "no longer well suited to the current policy environment."
He didn't submit a dot of his own.
This is worth sitting with. The new chairman of the Federal Reserve, in his first meeting, chose not to register a view on where rates should go. The committee — his committee — produced a hawkish dot plot without him. Either Warsh is more hawkish than the median and didn't want to pull the distribution further right in week one, or he genuinely hasn't made up his mind and is watching the data. Neither interpretation is bullish for the duration trade. One signals a chairman who is running ahead of the dots in private. The other signals a committee that is already hawkish independent of whatever its nominal leader believes.
Note also: Jerome Powell is still on the Board of Governors. He is a voting member of the FOMC. The prior chair — the man Warsh replaced, the man Trump spent two years publicly berating into submission — is sitting at the same table. That is its own kind of institutional complexity that nobody in the "Warsh dovish" camp adequately priced.
Now. Before we all agree to blame the Iran war and call it force majeure on the thesis, let's be precise about what the data actually says. CPI printed 4.2% year-over-year in May. The PCE number that drops this week will be the real test — consensus is somewhere around 3.6%, which happens to be exactly the Fed's own revised forecast. If it comes in hotter, September hike odds move above 67%, which is where they already spiked mid-Thursday before the Iran ceasefire news gave the equity market something to rally on. The brief afternoon recovery on Thursday was oil-driven relief, not a recalibration of the rates picture.
The deeper problem is the AI productivity thesis itself, which Warsh cannot deploy even if he wants to. Apollo's Torsten Slok put it cleanly in a note last week: the data center buildout is not sensitive to higher interest rates because FOMO among hyperscalers is not a rate-sensitive emotion. Meta, Microsoft, Alphabet — they are spending on AI infrastructure at a pace that implies the cost of capital is largely irrelevant to the decision. Fine. But that spending is currently inflationary. It is adding demand for chips, copper, and energy in quantities that overwhelm any near-term productivity offset. The AI disinflation story is a 2028 story. Maybe 2029. The AI inflation story is right now, and it is showing up in the numbers Warsh is legally required to address.
The market-implied fed funds rate for May 2031 now sits at 4.78%. Five hikes over five years, embedded in the curve. That is not a temporary repricing. That is a structural reassessment of what the terminal rate looks like in a world where AI capex is persistently inflationary in its construction phase, where energy prices remain geopolitically unstable, and where the new Fed chair has made clear — in forty minutes of unambiguous press conference language — that he intends to be orthodox on inflation regardless of what anyone in the West Wing was expecting.
There is still a path where Warsh gets what he originally sold. Oil holds below $85. PCE begins to moderate in Q3. The Hormuz deal stabilizes, energy supply recovers, and by November the inflation prints are trending toward 3% on a trajectory that justifies a hold and eventually a cut. The AI disinflationary story doesn't need to arrive by then — it just needs to look credible. In that scenario, the 2-year at 4.2% is the entry point, not the exit.
But that path requires a Brent price that is currently being negotiated in Geneva and Beirut simultaneously, a PCE print this Friday that cooperates, and a Fed chair who turns out to have been performing hawkishness rather than meaning it.
Three variables. All three need to break the right way.
We built a trade on one of them.
Circulate internally. Do not forward.