AI Megacaps Lost Their Grip on Oct 7. 5.36% Yields and $100 Brent Tell You Where the Money Actually Went

The FOMC minutes from the Sept 16 meeting, released Oct 7 at 2 p.m. ET, confirmed what dot-plot watchers had already inferred: Chair Kevin Warsh's committee voted unanimously 12-0 to raise the funds rate 25 bps to 3.75%–4.00%, and 16 of 18 participants placed the end-2026 rate midpoint at 4.125% or 4.375% — implying one more hike is on the table. The minutes said most officials see 'another rate hike by year-end' as appropriate if inflation doesn't cool.

That gap — hawkish minutes, dovish-priced October meeting — is what drove the bond rout. The 10-year closed at 5.276%, just below the 5.36% intraday peak, and the 30-year hit 5.73% intraday, the highest since June 2002. The 10-year has now risen 1.30 percentage points since the Iran conflict began, and the 30-year is up from 5.00% on Sept 15 to 5.73% on Oct 7. Long duration is being repriced, and the AI capex trade — which is itself a long-duration bet on cash flows 5–10 years out — sits directly in the line of fire.

Rates weren't the only pressure. Tanker attacks in the Strait of Hormuz hit their highest weekly pace since the start of the Iran war, and Iran accelerated strikes on shipping in the days before the Fed minutes. The result: Brent settled at $100.46, up roughly 9% on the day and more than 84% year-to-date. WTI closed at $88.28, off 1.16%, but Brent-WTI widened as U.S. shale supply failed to absorb the geopolitical premium.

OPEC+ kept November production targets steady, Saudi Aramco cut official selling prices for Asia (the largest discount since 2020), and the IEA moved to accelerate the release of 100 million barrels from strategic reserves. The supply response is real, but the market is still pricing for a sustained geopolitical risk premium rather than a quick return to $80.

The dollar is the cleanest expression of the rotation. The DXY rose 0.3% to 102.23, near a 17-month high. Gold fell for the first time in days — spot dropped to $4,146.70 and futures for October delivery settled down 1.34% to $4,130.80. The mechanics are textbook: a stronger dollar and higher real yields make non-yielding bullion less attractive in the very short term, even as the long-term thesis for gold as a sovereign-debt hedge strengthens.

On the S&P 500, eight of eleven sectors closed red and the winners told a defensive story. Health Care led at +1.06% to +1.09% — Moderna added roughly 5% and Eli Lilly gained nearly 3% — with Utilities flat-to-higher as the high-yielding sector caught a bid. Energy posted the day's best single stock, with NRG Energy up 4.85%. Industrials were the worst sector at -2.14% to -2.24%, with Caterpillar falling 5.77% to extend a 24% drawdown from its all-time high.

The energy side was the polar opposite. Exxon Mobil closed at $164.06 (-0.26%) and Chevron at $205.19 (-1.15%) — modest declines, but in the context of an index down 0.22% the relative strength is the story. Bank of America maintained a Buy on Chevron, citing 'better exposure to higher oil prices' than Exxon through the Hormuz risk window. Both names remain up 40%+ year-on-year, with XOM up 44.3% over the trailing year.

The most important data point in the AI tape on Oct 7 came out of Seoul, not Cupertino. Samsung Electronics flagged Q3 operating profit of 17.04 trillion won — roughly $80.2 billion at then-current FX — up 783% year-on-year and the largest single quarter in the history of any technology company. The LSEG SmartEstimate had been 10.61 trillion won. HBM, DRAM, and NAND all contributed, with Samsung's DRAM share reportedly back to 39% of the global market.

U.S.-listed Micron Technology — Samsung's direct DRAM and HBM peer — rose 4.1% to $1,088 on the read-through, lifting HBM-cycle exposure without the Korean stock's tax/fx overhang. SK Hynix fell 2.36% on profit-taking (it is still up meaningfully year-to-date and faces currency headwinds into its Oct 27 Q3 report), but the underlying message was the same: HBM scarcity is the cleanest AI trade right now, and pricing is passing through.

The AI-equipment side told a different story. TSMC ADR fell 2.09% to $472.20 despite its $2.5 trillion market cap and 50% gross margin. The SpaceX headline crystallized the risk: Bloomberg reported the company is in talks to raise $40 billion — split $10 billion in bank loans and $30 billion in bonds — to buy NVIDIA chips. That is one of the largest AI-debt financings ever, and the bond leg adds directly to the Treasury supply that's pushing the long end higher. The transmission: AI capex is now funding itself by issuing debt, that debt hits the curve, the curve pushes yields up, yields discount AI's terminal value, and AI mega-caps correct. It is a self-reinforcing loop, and Oct 7 is the first time the market clearly saw it.

Bitcoin fell from the $86,000 area to $83,065 in the 15 minutes after the 2 p.m. ET Fed minutes release, and ended the Oct 7 session near $83,300. By Oct 8 morning it was at $82,861 — a 1.71% additional drop. Total crypto market cap slid 3% to $2.85 trillion, with roughly $88 billion erased. Long-side liquidations on Oct 7 reached $143 million, with the largest single block at 06:53 UTC.

The decoupling from equity flows is what stands out. Spot Bitcoin ETFs absorbed $118.8 million of net inflows on Oct 6 — the day before the selloff — with BlackRock's iShares Bitcoin Trust alone pulling in $122 million. By Oct 7 the same fund held roughly 800,000 BTC and net assets of $9.62 billion; it traded down 2.64% on the day, tracking spot, but did not see outflows. The message: institutional buyers are still adding on the way down, even as leveraged retail gets forced out.

Ethereum took the bigger hit than Bitcoin on a percentage basis: ETH traded at $2,562 at 10:32 a.m. ET on Oct 7, down 5.5% over 24 hours, and spot ETH ETFs logged their sixth consecutive day of outflows — a $201.9 million exit on Oct 6 was the largest single-day outflow in three weeks. The divergence is consistent with the prior pattern: when the dollar bid and rate cycle dominate, Bitcoin's scarcity narrative holds better than Ethereum's DeFi-utility narrative.

The Oct 7 session is the cleanest single-day expression yet of the regime the market has been sliding into since the Iran conflict began: a dollar bid, a long-end yield blowout, and a rotation out of long-duration AI capex into short-duration dollar hedges, energy, and gold. Three things will determine whether that regime breaks in the back half of October:

Until one of those breaks, the cross-asset posture is clear: own the dollar, own the energy majors, own HBM/pricing-leverage memory names, and own the gold miners as a hedge against the 30-year staying above 5.5%. The trade to fade is the AI equipment complex funded by debt issuance — SpaceX's $40 billion is the headline, but it is not the only one. Each additional billion of AI debt adds incremental supply to a long end that is already at 2002 levels, and the AI trade has no margin for further yield backup.