10-year Treasury yield: the long end moved. The Fed did not.

The 10-year Treasury yield is the search term. The story is the 30-year. Policy is still 3.50–3.75%. If your model is “yields up = the Fed hiked,” you will misread every headline between now and Jackson Hole.

Price of a bond and yield move opposite. A sell-off in long Treasuries is higher yields. Nobody has to wait for Kevin Warsh to vote for that. The market did it in the cash curve.

The print

Board of Governors H.15, release date Aug. 18, 2026 — constant-maturity Treasury yields, percent:

• Aug. 17: 2-year 4.19%, 10-year 4.72%, 30-year 5.31%
• Aug. 14: 10-year 4.68%, 30-year 5.25%
• Aug. 13: 10-year 4.63%, 30-year 5.21% (the week’s local dip)
• Effective federal funds all five days: 3.63%. Primary credit 3.75%. Bank prime 6.75%.
• 10-year TIPS (inflation-indexed) Aug. 17: 2.44%. Rough breakeven: 4.72 – 2.44 = 2.28 percentage points. That is the market’s average CPI path over a decade, not next month’s print.

Source: Federal Reserve H.15, Aug. 18, 2026. Intraday Tuesday (Aug. 18) on screens can print a tick above or below those closes. Use H.15 for the official CMT series; use the cash 4:00 p.m. quote if you are marking a book.

2s10s on Aug. 17: 4.72 – 4.19 = 53 basis points. CNBC, using FactSet, had 2s10s wider by about 29 bp since June 24. Same shape: the Fed holds the front; the back end does the talking.

A bear steepener, in English

Four cartoon curves:

1. Bull steepener — short rates fall more than long (cuts coming, recession scare).
2. Bull flattener — long rates fall more (growth scare, duration bid).
3. Bear flattener — short rates rip (hike cycle). The 2-year does the damage.
4. Bear steepener — long rates rip while the front sits. Term premium, supply, inflation-at-the-horizon. That is this tape.

July 29 FOMC: 9–3 hold at 3.50–3.75%. Three presidents wanted +25 bp. The 2-year around 4.2% already embeds some extra tightness versus 3.63% effective funds. The 30-year at 5.31% is not “the Fed.” It is duration supply plus a fatter inflation/fiscal premium. We already walked the vote and the Wyoming calendar: Jackson Hole 2026.

If Wednesday’s minutes (July 28–29 meeting, due three weeks later: Aug. 19) sound hawkish, the 2-year can catch up. That would flatten from the front, not from a 30-year rally. Different trade.

Mortgages and Main Street

Freddie Mac Primary Mortgage Market Survey, week of Aug. 13: 30-year fixed 6.67%, 15-year 5.96%. A year earlier the 30-year was 6.58%. Source: Freddie Mac PMMS. FRED series MORTGAGE30US matches that 6.67%.

Mortgage quotes on a Tuesday can sit closer to 6.75% while the weekly survey is still 6.67%. Both can be true. The PMMS is applications through last Thursday. The 10-year jumped into Monday–Tuesday. Lenders reprice daily; Freddie averages weekly.

The wedge (mortgage minus 10-year) is originator cost, G-fees, servicing, and credit. It is why “the Fed held” and “my refi quote went up” coexist. Existing 3% coupons do not reprice. New purchases and cash-out refis do. Housing turnover stays the valve.

A HYSA still tracks the front of the curve and bank deposit betas, not the 30-year. I-bond composites track CPI-U with a lag, not CMT 10s: I bonds vs high-yield savings.

Why the long end

Three forces that do not require a hike:

Supply. Coupons keep coming. Corporate AI/data-center issuance competes for the same duration buyers. Heavy IG supply does not need a conspiracy; it needs a bid. When the bid steps back, yields gap.

Inflation that is not “core.” July CPI-U +3.4% over 12 months; energy +14.7% year over year. The FOMC statement already named energy supply shocks and the Middle East. Long bonds care about whether $90-ish Brent is a one-month spike or a year of gasoline in the CPI. COLA 2027 uses CPI-W Q3, a different formula, same BLS building: Social Security COLA 2027.

Fiscal. Interest on a ~$40 trillion stock of debt is a line item that grows when the 10-year and 30-year do. That is mechanical. “Bond vigilantes” is a magazine word for “duration buyers asked for more yield.” The Fed does not control the coupon calendar.

TIPS 10-year at 2.44% real means if your forecast of 10-year average inflation is 3%, the 4.72% nominal 10-year is not a gift. If your forecast is 2%, it is. That arithmetic is the whole argument. Screenshots of “5%” without the real yield are noise.

Stocks are a duration trade

A higher risk-free discount rate cuts the present value of cash that arrives in 2034. Unprofitable growth and long-duration AI stories feel it first. A bank with floating assets can like a steeper curve. A hyperscaler funding a campus with new bonds does not.

Tuesday tape (wires): Nasdaq harder than the Dow while the 30-year tagged a 19-year high. That is textbook duration, plus oil. It is not proof the S&P “broke.” Three years of equity returns can absorb a 20 bp backup until they cannot. The Mag 7 / GPU / optics stack still needs a cost of capital: AI semiconductor and optical stocks.

Buybacks do not repeal duration math. A company retiring stock with cash it could have left in T-bills at 4% is a different NPV when the 10-year is 4.7% than when it was 3.8%: stock buybacks.

This week on the calendar

1. Aug. 19, 2 p.m. ET — minutes of the July 28–29 FOMC. Three-week lag is statutory habit, not a leak. Watch whether the 9–3 was a near-tie in the room or three isolated hawks.
2. Aug. 27–29 — Jackson Hole. Payments theme. Do not expect a 25 bp announcement from a lodge.
3. Sept. 11 — August CPI.
4. Sept. 15–16 — FOMC plus SEP dots. That is the vote.

If the 30-year keeps making 2007 highs while funds stay 3.63% effective, the market is pricing term premium, not a shadow hike. If the 2-year rips toward 4.6% after the minutes, then you are back in a hike narrative. Those are different sentences. The 10-year is just the one everyone types into Google.

Education, not a recommendation to buy or sell bonds, bond funds, or stocks. Yields on H.15 are constant-maturity series. Your brokerage quote is a specific CUSIP. They will not match to the basis point.