The 5% Wall Held: How a 4.98% Touch, a Hot CPI, and a Priced-In Fed Hike Gave Bessent His Biggest Reprieve Yet

For three trading days the bond market had been telegraphing that5% was next. The 10-year yield climbed from 4.78% on September 4 to 4.83% on September 9, then ripped to 4.95% on September 10 after the August PPI print came in at +5.4% year-over-year — a tenth above expectations and the fastest wholesale-inflation pace this year. The 30-year broke to 5.37%, the highest closing level since 2007. By the New York open on September 11, with August CPI running +3.4% year-over-year for the second straight month and core inflation refusing to roll over, the 10-year touched 4.979% intraday — the highest print since late 2023.

Then it stalled. The yield reversed on the day, closing around 4.93%, off 1 basis point. Reuters framed the move as a \"reprieve for Bessent\"; the 5% line, for now, has held.

Treasury Secretary Scott Bessent spent August escalating the intervention toolkit. On August 19 he announced that the Treasury would double the maximum size of its long-bond repurchase operations — lifting the per-auction cap from $2 billion to at least $4 billion, with the operational program now running around $6 billion in the 10-to-20-year sector. The move was framed officially as a liquidity operation to thin out off-the-run inventory, but the timing — yields had hit 4.74% in the wake of the Iran war and the 30-year had just cleared 5.33% on August 18, the highest level since June 2007 — made the yield-management intent unmistakable.

The same package introduced what Yardeni has labeled the \"Bessent Twist\": finance larger long-bond buybacks by issuing more Treasury bills, leaning on the $31.8 trillion Treasury market's $5.5 trillion of long bonds to compress term premium without changing headline issuance. Bessent has also signaled willingness to draw down the Treasury General Account to fund additional purchases, and to coordinate across the curve — including public support for the yen — to slow the rise in real yields. The September 11 intraday reversal at 4.98% is the first clean test of whether this toolkit can hold the 5% line.

If the 5% line holds, the marginal cost of capital for long-duration cash flows stabilizes. The \"big five\" hyperscalers are forecast to spend roughly $602 billion on AI capex in 2026 — a 73% jump from 2025, per CreditSights — with broader analyst estimates closer to $700–$775 billion once Microsoft, Alphabet, Amazon, Meta Platforms, and Oracle are aggregated. NVIDIA sits at the top of the supplier stack. A 10-year pinned near 4.9% instead of clearing5% shaves meaningful basis points off the discount rate applied to multi-year AI revenue ramps — the present value of a dollar of revenue deferred to 2030 is roughly 5% higher at 4.9% than at 5.5%.

The second-order effect is at least as important. iShares' Fall 2026 Investment Directions explicitly identifies AI capex as a structural force pushing longer-term yields higher: more corporate bond supply to fund data-center buildouts competes with Treasuries for buyer demand. If the 10-year caps at 5%, the implied incremental cost of financing another $100B of AI capex is materially smaller. Hyperscaler free-cash-flow profiles — which several firms have flagged as vulnerable under higher discount rates — get a marginal reprieve.

On the other side of the duration trade, regulated utilities are the most rate-sensitive equity sector. NextEra Energy and its peers trade on long-duration contracted cash flows that re-rate mechanically with the 10-year; a 5% ceiling limits the multiple compression that began when yields pushed through 4.6% in May.

Agency mortgage REITs are the listed equity pocket with the most mechanical exposure to the 10-year and30-year. AGNC Investment reported a Q2 2026 tangible net book value of $8.58 per common share; Annaly Capital Management reported a Q2 2026 book value of $20.15 per share, with adjusted EPS of $0.79 and a quarterly dividend raised to $0.75. Both run 5–8x leverage against agency MBS, funding them with repo and short-term debt whose cost is benchmarked to short rates while the assets ride the long end. A 10-year that can't break 5% caps the mark-to-market drawdown on their portfolios — and the 30-year is the more important number: 5.37% on September 10 vs. 4.63% post-buyback on August 19 shows how violent the swing has been.

The 30-year fixed mortgage rate hit6.76% in the week ending September 11, per industry trackers, with at least one daily survey printing 6.85%. Housing starts fell 12.4% in July to a 1.24 million annualized pace. Every basis point of 30-year yield that gets capped at 5.5% instead of running to 6% is real money in origination volumes, real estate services revenue, and bank mortgage banking fees.

The most striking number in the September tape isn't the 4.979% intraday high — it's the 80% Polymarket odds on a 25 basis point Fed rate hike at the September 16–17 FOMC meeting. The market is now pricing the Fed to raise into a5% ten-year, the first time the expected policy direction has been unambiguously hawkish since the June 2026 dot plot signaled a divided committee with nine of eighteen members still preferring at least one more hike this year.

If the Fed delivers the hike and the 10-year still can't clear 5%, the message is that the term premium — not the expected fed funds path — is the binding constraint. That is the regime boundary the brief is pointing at: structural demand at5% from pension rebalancing, foreign reserve managers, and now Treasury's own buyback operation is absorbing the entire flow. It also means equity duration trades get a reprieve without requiring the Fed to pivot dovish.

Over days to weeks, the September 16–17 FOMC decision is the binary catalyst. A 25 bp hike delivered against a 10-year pinned near 4.95% would confirm the regime read; a pause would force a violent rally in the long end and a relief rally in NVIDIA and the hyperscalers. The September 1610-year Treasury auction will be the most-watched refunding of the cycle — a tail above5% would crack the wall.

Over quarters, the calculus shifts to net issuance. Treasury has roughly $1.2 trillion of coupons and bills to roll over in Q4 2026, with the 10-year sector carrying the heaviest burden. The buyback program is roughly $6 billion per operation — a rounding error against the $5.5 trillion of long bonds outstanding, but a useful signaling tool. The structural question is whether Bessent's toolkit can hold5% through a refunding cycle that will also be competing with $200 billion-plus of incremental AI capex bond issuance.

Over one to three years, the question is whether the term premium stays capped. If the Fed cuts into 2027 (the June 2026 dot plot had the median dot heading toward 3%), a 10-year pinned near 5% would imply a structurally steeper curve and a meaningful easing of financial conditions — bullish for housing, mortgage REITs, and the AI capex complex, but a warning sign for the dollar and for any fiscal path that relies on cheap long-term financing.