The Disruption Leg: When AI Comes for the Financial Middle-Men
The funding-stack thread on this blog — China's Treasury unwind to 2008 lows, Nscale's documented vendor guarantees, SoftBank's junk-rated ladder rung — has tracked one transmission channel: AI straining its own credit. Tuesday's tape opened a second channel that the four-pillar framework had underweighted: AI products destroying the revenue pools of the intermediaries who finance the buildout.
The tape
Meta's Muse consumer assistant did not launch a brokerage product. It didn't need to. The market repriced the middle-men anyway:
- Schwab -6% Tuesday, lowest since early July, -8% since the Muse announcement; options volume ran 5x average with nearly 2x puts vs calls, and the most-bought contract was the January 90-strike put — a three-month structural bet on another 13% lower, not a day-trade.
- LPL Financial -7%, Raymond James -3.5%, Interactive Brokers -1%.
- XLF -2% on the day, -6% from its September highs, now printing RSI 29.9 below every daily moving average.
The tell is inside the sector: Robinhood rose to a YTD high on the same day, with 80% of $95M in options premium on the call side. The market is not short 'brokers.' It is long displacement within the group — asset-light platforms with AI agents already deployed (Robinhood's May launch) against fee-heavy incumbents monetizing what Piper Sandler calls 'the frictions legacy intermediaries have historically monetized.' Schwab's low-yielding sweep-cash economics sit, in that framing, 'squarely in the crosshairs.'
Why this belongs in the framework
Pillar 1 (bear steepening) and Pillar 4 (AI capex durability) both assumed AI's costs land on credit markets and its benefits land on hyperscaler earnings. The Muse tape adds a third leg: AI's costs also land on financial-sector earnings — billing, reporting, trading, advisory fee pools. At a 10Y near 5%, XLF now fights a two-front war: funding costs rising AND the fee pools that justify its valuation under existential question. Financials were this framework's core steepener expression. The steepener itself is intact (2s10s +19bps, 30Y 5.31%), but the equity expression is being repriced for a risk that did not exist in August.
The honest counter-case
This has happened before. In February, Altruist's 'Hazel' tax-planning agent knocked the same group down; Schwab recovered within months, LPL never did. Muse states it won't trade for users or give financial advice. Schwab is counter-positioning with Anthropic (Claude for Financial Advisors) and shipped its own AI research tool in May. This can be February 2.0 — a rotation head-fake.
Three things argue it is not. First, distribution: Muse ships to Meta-scale user bases, not a niche prosumer platform. Second, the positioning is structural — a January-dated put, not Friday gamma. Third, the same week the neocloud debtor layer cracked too: CoreWeave, -40% from May highs, launched a $3B convertible (due 2033, upsizable $500M) plus a program to sell up to 35M shares — equity-dilutive funding at the AI infrastructure layer. Meanwhile the compute layer (NVDA, above every daily EMA, MACD bull) and the index layer (QQQ record close) print strength.
That is dispersion, not uniformity: strain at the debtor layer, disruption at the intermediary layer, strength at the compute layer. Dispersion regimes are stock-picking regimes. A VIX at 14.3 pricing index-level calm while single names move 6-7% on product announcements is an inconsistency that usually resolves upward in vol, not downward.
Positioning consequences
- The XLF/financials long expression is under review — not exited. The curve thesis survives through duration and the steepener, but sector-equity exposure now carries an idiosyncratic disruption risk absent from the August framework.
- SCHW joins the watchlist as a short candidate: below the 100.32 Camarilla S2 with follow-through = trigger; above the EMA20 near 107.6 = invalidation. No entry — the Trump-Xi summit binary and post-FOMC freeze still govern.
- The falsifier is clean: if SCHW stabilizes above 100 this week and XLF reclaims its EMA20, this was February redux. If the January 90-put starts paying, the disruption leg is real, and 'AI kills the middle-man' becomes a 2027 earnings model, not a headline.
The AI trade's next victims were never going to be only the companies that overborrowed. They will include the ones that stood in the middle of the fees.