TL;DR
The Fed's hawkish September hike has hardened into a priced-in October follow-through — CME odds briefly hit 73% after Governor Barr's remarks — while the 10-year pushed above 5% on an unexpected driver: the AI buildout itself, now issuing bonds at a pace that's moving the Treasury market. Equities shrugged anyway, carried almost entirely by megacap tech, while the rotation into small caps and value that defined mid-2026 went into reverse. Meanwhile, the oil risk premium that briefly deflated snapped right back when Trump rejected Iran's Hormuz proposal.
The Fed Is Hiking Into Strength, and the Market Has Priced the Second Hike
The October 27–28 decision has moved from open question to near-consensus, and the data keeps arguing for it.
"In my base case, further policy adjustments are likely to be needed to ensure inflation comes down to target in a timely fashion." — Kevin Warsh's Fed: October Hike Odds Surge Past 70%
(via CNBC)
Barr's September 23 remarks landed the same day S&P Global's flash PMIs showed the economy accelerating into the tightening — services at a 59-month high, with input costs jumping at the steepest rate in four years. CME FedWatch odds of an October hike hit 73% on the news and stood near 66% by Sunday, with the dot plot showing 16 of 18 participants expecting another increase this year and no cuts until 2028 (Kevin Warsh's Fed). Lance Roberts frames the asymmetry plainly: a committee this confident about growth "isn't positioned to deliver 'rate cuts' quickly if earnings disappoint." The week's lock-in events are August PCE Wednesday and September payrolls Friday — with a wrinkle: the BEA is changing PCE methodology, and RBC warns against misreading a soft print as disinflation.
What to watch: A hot-hot PCE/payrolls combination, which would all but guarantee the October hike against a 10-year already above 5%.
The AI Buildout Is Now a Bond-Market Force — and It's Lifting Your Mortgage Rate
The most consequential new driver of yields isn't the deficit or the Fed; it's hyperscaler debt issuance.
"...the 30-year Treasury briefly hit 5.501% on Sept 24 — first time since June 2004 — and the 10-year settled near 5.17%, highest since 2007." — AI Capex Returns: The $570B AI Bond Wave
(via TechTimes)
Morgan Stanley estimates AI-linked bond issuance near $570 billion in 2026, with hyperscaler capex running close to 100% of operating cash flows and borrowing up from 9% to 32% of capex. Barclays notes the six largest tech companies now exceed the six largest banks in investment-grade index duration exposure — meaning target-date funds must absorb every new AI bond proportionally (AI Capex Returns). This changes the risk map for equity investors: BofA's fund-manager survey now ranks a disorderly yield rise and an AI bubble as the two largest market risks, and both stem from the same trade. Madison Investments' Mike Sanders captures the shift — the rise in yields "can no longer be attributed simply to concerns over the deficit."
What to watch: Whether the 10-year sustains trading above 5.5%, which would compress the already razor-thin equity risk premium and pressure AI-adjacent valuations.
Stocks Are Ignoring the Bond Market — With an 8-Basis-Point Cushion
The equity-bond divergence that snapped in June has reopened at a higher level, and the valuation math is now uncomfortably tight.
"Stocks have ignored the bond market for two weeks. Wednesday's PCE report and Friday's payrolls will test whether they can keep doing so." — Equity-Bond Divergence, Round Two
+1 (via Real Investment Advice)
FactSet's 19.1x forward P/E implies an earnings yield of roughly 5.2% against a 5.11% 10-year — investors are paid about 8 basis points to own stocks over Treasuries. Goldman's speed rule says stocks struggle once the 10-year moves ~30bp in two weeks; it's up 28bp since September 9. Yet the S&P 500 gained 1.2% to 7,744.64, sitting just 0.7% below its August record (Equity-Bond Divergence+1). The internals tell the fragility story: the equal-weight index fell 1.1% while the cap-weighted index rose, seven of eleven sectors declined, and technology (+3.6%) carried everything. Ruchir Sharma predicts the AI bubble pops when the 10-year decisively exceeds 5% — a line the market is currently dancing on.
What to watch: The 7,800 battleground on the S&P — a close above confirms the breakout; a failure with breadth this narrow keeps the index rangebound.
The Great Rotation Didn't Just Stall — It Reversed
The mid-2026 rotation into small caps and value was always a rate trade, and 5% yields killed it.
"Growth, Speculative Technology, and Megacaps now extremely overbought, while Value, Low Beta and Dividend Yield... now the most oversold. A risk-off rotation seems highly probable." — The "Great Rotation" Stalls
+1 (via Real Investment Advice)
The week's damage map was the mirror image of the rotation thesis: utilities fell 4.3%, energy dropped 3.8%, real estate lost 2.2%, and the Russell 2000 fell 1.8% on Wednesday's yield spike alone (The "Great Rotation" Stalls+1). RIA's advice cuts against chasing either side — rebalance tech winners back to target weights, because "too much of this week's gain came from a single sector." The counterweight worth holding: the beaten-down rate-sensitive sectors are exactly where any yield pullback would spark a violent rotation back into.
What to watch: Friday's payrolls print — with CTAs holding their largest duration shorts since April, a soft number could force yields down and restart the value trade.
Micron's Print Is the AI Trade's Cleanest Test — and the Bear Case Is Building Underneath
Consensus has moved above management's own guidance, which means a beat-and-raise is fully priced.
"If AI capex remains durable, MU's earnings power and valuation remain compelling; if AI spending weakens, customer purchases and future capacity additions will both adjust." — Micron Into the $50B Quarter
LSEG consensus sits at $50.9B revenue and $31.49 EPS versus a $50B ± $1B guide, with the real event being the take-or-pay SCA count — Rosenblatt expects SCAs to account for up to 40% of bits produced over the next 3–5 years (Micron Into the $50B Quarter+1). But the bear signals are accumulating: TrendForce sees DRAM price increases decelerating sharply, IDC now expects smartphones to contract ~17% on memory costs, and AI inference pricing is deflating fast — OpenAI cut developer API pricing roughly in half. At ~7x forward earnings, the market is paying for cycle durability, not level.
What to watch: The September 30 margin guide — anything short of strength invites the "memory is still cyclical" repricing, on the same day as PCE.
Oil's Risk Premium Came Back in a Week
Last digest noted the deflating war premium; that lasted about seven days.
"...Brent crude rebounded more than 3% to $107.75... after US President Donald Trump rejected a peace deal from Iran to resolve their conflict and reopen the Strait of Hormuz." — Oil's Risk Premium Returns
+3 (via Reuters)
Diplomacy is stalled, not dead — Trump told Axios he expects negotiators to keep talking — but the physical picture complicates the narrative: Mideast exports actually hit their highest level since the war began in February, per Kpler (Oil's Risk Premium Returns+3). The equity link is direct: University of Michigan year-ahead inflation expectations rose to 4.6%, and the Fed "cannot reopen the Strait of Hormuz" — the inflationary impulse is supply-side, not demand-driven.
What to watch: Whether Brent sustains above $120 resistance — the level analysts peg as the line between pullback and breakout.
SpaceX Is Quietly Passing Its Supply Tests
The lockup overhang that Motley Fool and others flagged as an existential supply problem keeps getting absorbed. The September 24 tranche — up to 328.4 million shares, roughly 7% of shares outstanding — passed without a collapse, leaving SPCX around $148.68, roughly 10% above its $135 IPO price (SpaceX: Second Lockup Tranche Absorbed). The pre-unlock wobble came from Gwynne Shotwell's first major post-IPO sale filing (~$52 million under a 10b5-1 plan), not from the unlock itself. The structural backstop: Musk is locked up until June 2027, and the Nasdaq-100 index bid partially offsets insider supply.
What to watch: The December lockup expiration — the largest remaining scheduled release of 2026, and the last big overhang test.
What surprised us
- The long end is being driven by the AI trade, not the Fed. The 30-year hit a 22-year high not because of deficits or hawkish dots but because hyperscalers are issuing hundreds of billions in bonds — and the Treasury Secretary doubled buybacks to cushion it, with limited effect (AI Capex Returns
- The rotation died by its own mechanism. The small-cap/value trade thrived on expectations of one-and-done tightening; the Warsh Fed's open-ended hike path is precisely what killed it. Rate expectations, not fundamentals, are the rotation's on/off switch (The "Great Rotation" Stalls
- Memory demand is destroying its own downstream customers. IDC sees the smartphone market contracting ~17% because memory costs have pushed entry-level devices out of viable price points — the supercycle is cannibalizing non-AI demand (Micron Into the $50B Quarter
- Two AI sentiment shocks were fully absorbed. An Amodei-led AI-safety selloff and an OpenAI pause report both knocked chips before a rebound restored roughly $3 trillion of market value — the complex is currently buying every dip (AI Capex Returns