The Fed hiked rates for the first time in three years, yet bank stocks were the worst performer of the week. Shouldn't higher rates mean banks make more money?
2026-W38
Rates rose, but not the rate banks earn on
Rates didn't move in one direction this week. The 3-month yield (^IRX) rose +1.79% to 3.98%, while the 30-year (^TYX) actually fell -0.37% to close at 5.33%. External data shows the same picture. As of September 17, the 1-year yield rose 12bp to 4.40% and the 2-year rose 11bp to 4.67%, while the 30-year fell 8bp to 5.29% (The Curiosity Vine, 2026-09-17). The short end rose while the long end fell, flattening the yield curve — and that single fact explains the bottom of this week's sector heatmap.
A bank's earnings structure is simple. What it pays depositors tracks short-term rates; what it earns on loans and held bonds tracks long-term rates. This week only the pay side rose while the earn side fell. It's a sharp contrast with 2025, one of the best years on record for banks, precisely because the opposite was happening — short rates falling and the curve steepening, widening net interest margins (Angel Oak Capital, 2026 Financials Outlook). XLF's -2.43% is a result of the shape of the hike, not the hike itself.
"Banks make money on the spread between what they pay to borrow (short-term deposit rates) and what they charge to lend (long-term loan rates). A steepening curve widens that spread; a flattening curve narrows it." — MarketWise, analysis of bank stocks and the yield curve (2026)
Utilities can't be explained by rates alone
Here many investors draw the wrong conclusion — that because three rate-sensitive sectors landed at the bottom together, there must be a single cause. But the week's worst performer, XLU at -3.04%, doesn't fit that story. Utilities are sensitive to long-term rates, and the 30-year actually fell this week. Rates alone can't explain why utilities fell harder than financials.
The second driver is a re-rating of AI power demand. Utilities have re-rated this year from defensive stocks to "growth stocks selling electricity to AI data centers," earning a premium along the way. That premium is now facing headwinds. Goldman Sachs estimates that AI infrastructure buildout will push up electricity rates 6% in 2026-2027 and another 3% by 2028 (Goldman Sachs estimate, 2026), and rate hikes translate directly into regulatory risk. Indeed, the dominant theme of utilities' Q1 2026 earnings season wasn't growth but "affordability," with the industry split over data-center contracts (Utility Dive, 2026). Some have also noted that a $7 trillion buildout plan isn't backed by confirmed AI demand (Fortune, 2026-07-26).
So even though the three sectors sit together at the bottom, the character differs. REITs (XLRE -2.05%) are a pure casualty of the 10-year sticking near 5%; financials are a casualty of the curve's shape; and utilities are a casualty not of rates but of the AI-theme unwind. That means the recovery paths differ too.
Next week's follow-through depends on the shape of the curve
| Next week's curve | Condition | Sector implication |
|---|---|---|
| Flattening persists | Short end rises, 30-year stalls near 5.3% | Financials remain weak, long bond ETFs stay positive |
| Re-steepens | 30-year returns above the weekly high of 5.39% | Financials rebound, REITs and long bonds pressured again |
| Rises across the board | 10-year retests and holds above 5% | All three sectors fall together, a stagflation-type combination |
The third path is the most dangerous. In a slowing-growth environment where rates rise and the curve flattens, warnings already exist that declining loan volumes, rising credit costs, and mark-to-market losses on held bonds outweigh any limited NIM improvement (Angel Oak Capital, 2026). Layer on the fact that private-credit default rates hit a record high on September 16, and it's hard to view bank weakness purely as "temporarily pressed by the curve."
What should investors do
- Financials (XLF): Watch the spread between the 30-year and 2-year, not the policy rate, as the key confirmation gauge. There's no case for new entries until that spread widens again. Earnings from KB Home, Cintas, Paychex, and Costco running from September 22-25 offer a first read on loan demand, employment, and consumer resilience.
- Utilities (XLU): Approaching XLU on the premise that "it'll rally when rates fall" will never explain this week's losses. What's moving XLU right now is data-center power contract and rate-regulation news, not rate headlines.
- Long bonds (TLT +0.47%): Remember that the sole reason for this week's gain was the 30-year decline. If the 30-year turns back up, this is the first position that unwinds.