Everything About AI
NextFin News - The S&P 500 Financial index fell 2% on Tuesday and its bank gauge dropped 3%, as two separate threats to the sector's earnings model converged in a single session: investors began pricing artificial intelligence as a genuine competitor to fee-based wealth management, while the Treasury yield curve flattened to its tightest level in more than a year and a half after the Federal Reserve delivered its first rate hike in over three years. Charles Schwab led the decliners with a 6.1% loss, Ameriprise Financial fell 4.4%, and Raymond James shed 3.5% — on a day when the S&P 500 itself finished nearly flat and the Nasdaq Composite rose to a fresh record close.
The Setup: Two Headwinds, One Sector
The selectivity of the selloff is the first clue that this was not a simple risk-off day. The benchmark S&P 500 ended the session essentially unchanged, and the technology-heavy Nasdaq Composite advanced 122.18 points, or 0.5%, to 27,244.28, a new record. The pain was concentrated in the balance-sheet and fee-driven businesses that make up the financial complex. That divergence points to sector-specific drivers rather than a broad retreat from equities.
The first headwind arrived from Silicon Valley. Meta Platforms' new AI agent, Muse, overtook OpenAI's ChatGPT as the most downloaded free iPhone app in the United States, according to app-analytics firm Sensor Tower. The app, which launched on September 8 and is powered by Meta's Muse Spark family of models, logged roughly 730,000 downloads in its first five days. For portfolio managers who have spent years treating AI as a back-office productivity tool, a consumer-facing agent with Meta's distribution reaching into financial advice territory reframes the threat as a revenue-side problem.
The second headwind came from the bond market, where the spread between the 10-year and 2-year Treasury yields — the classic gauge of bank lending profitability — narrowed to an intraday low of 17.90 basis points before settling at 21 basis points. That is the flattest the curve has been since March 2025 and a steep drop from the 55.5 basis points recorded on August 18. The move follows the Federal Reserve's September 16 decision to raise its benchmark rate by a quarter percentage point to a target range of 3.75% to 4%, the first increase since July 2023, on a unanimous 12-0 vote.
The two forces attack the same line item from opposite directions: the AI agent compresses the fees banks and brokers can charge, while the flattening curve compresses the net interest margin they earn on deposits and loans. Together they explain why financials underperformed on a day when the broader market barely moved.
The Yield Curve: Why Flattening Hurts Banks More Than Headlines Suggest
The mechanism is straightforward but often misunderstood. Banks borrow short and lend long. They pay depositors rates that track the front end of the curve — the 2-year yield, which sits near 4.73% — and charge borrowers rates anchored to the long end, where the 10-year yield trades around 4.93%. The spread between those two is the raw material of net interest income. When that spread was 55.5 basis points in mid-August, a bank could cover its funding costs and still pocket a meaningful margin. At 21 basis points, the same balance sheet earns a fraction of what it did six weeks earlier.
The flattening is not an accident of positioning. It is the direct arithmetic of the Fed's September hike. The front end rose on the policy decision and on the central bank's own guidance: the September Summary of Economic Projections showed 16 of 18 FOMC participants expecting at least one more rate increase in 2026, and rate futures were pricing roughly a 90% probability of another quarter-point hike by year-end. When traders price in successive short-rate increases while the long end stays anchored by growth and inflation expectations, the curve flattens by construction.
"There's a tipping point between raising rates reflecting a strong economy and raising rates and having the effect of slowing the economy," said Rick Meckler, partner at Cherry Lane Investments.
That sentence captures the second-order risk. A steep curve signals that the market expects growth and inflation to persist. A flat curve signals that investors believe the Fed is about to over-tighten — that today's strong economy is the setup for tomorrow's slowdown.
Here is the distinction that matters for the cyclical-versus-structural call: this leg of the flattening is cyclical. It is driven by a repricing of the near-term policy path, and policy paths reverse. If inflation cools faster than the Fed's projections, or if the labor market shows cracks, the front end will fall and the curve will steepen again. The evidence for mean reversion is in the history of the last two decades: every post-crisis flattening episode tied to a hiking cycle eventually unwound once the hiking cycle ended. The 2s10s spread is a policy derivative, and policy derivatives revert when the policy cycle turns.
But there is a structural layer underneath. The long end has not risen in lockstep because the market is quietly questioning whether the AI-driven capital expenditure boom will deliver the productivity growth needed to justify today's valuations. If the long bond is right and the short end is wrong, the flattening is not just a cycle — it is a warning that the economy's nominal growth engine is weaker than the Fed assumes. That is the version of this story that does not revert on its own.
The AI Agent: From Cost Saver to Revenue Competitor
For most of the AI boom, financial institutions framed machine intelligence as a margin story: automate the call center, speed up underwriting, cut compliance headcount. The market rewarded that framing. What changed on Tuesday is that investors began pricing AI as a demand-side threat to the highest-margin line of business in finance — wealth management and advisory fees.
The catalyst was distribution, not capability. Meta's Muse did not need to prove it could out-pick a human portfolio manager. It needed to sit on a consumer's home screen, free, inside an app they already use for hours a day, and be good enough to answer "how much can I spend this month?" or "should I refinance?" Macrae Sykes, a portfolio manager at Gabelli Funds who invests in the financial industry, put it plainly:
"There's worries about AI disruption to traditional businesses. Clearly there's going to be some arguments both around the benefits and the disruption of AI tools and what it all means for the incumbents, but I think today is another sell-without-regard reaction."
That last clause is doing important work. Sykes is describing a reflexive move — a price adjustment that runs ahead of any actual revenue impact. And he is right that no AI agent has yet taken a meaningful share of assets under management. But the structural question is whether the moat around advisory relationships is as wide as the industry assumed. If a free agent can handle the routine questions that justify a 1% advisory fee, the fee itself comes under pressure even if the assets never leave.
This is the structural leg of the thesis. Unlike the yield curve, AI agent adoption does not mean-revert. Once consumers become accustomed to a free, always-available financial interface, the burden shifts to incumbents to prove why their human intermediation is worth paying for. The distribution advantage is real: Meta controls the attention layer where financial decisions begin, and it does not need to monetize the agent directly — it needs to monetize the data and the commerce that flows through it. That is a different business model from charging basis points on assets, and it is one that traditional brokers cannot easily replicate.
The counter-argument is equally strong and worth stating in full. Wealth management is a trust business built on fiduciary duty, regulatory oversight, and deep integration with tax, estate, and lending relationships. An app that answers questions is not a substitute for a relationship that holds $2 million of a family's net worth. Schwab, Ameriprise, and Raymond James are not pure fee-takers — they are custody platforms, lenders, and retirement plan administrators. The switching costs are high, the regulatory perimeter is thick, and the average client is older and more conservative than the early-adopter demographic that downloads new apps. On this read, Tuesday's 6% drop in Schwab is a sentiment overshoot, not a repricing of a durable earnings stream.
The falsifying signal that separates these two views is concrete: monthly net new asset flows at the major brokers. If Schwab, Ameriprise, and Raymond James report three consecutive months of net outflows from advisory accounts while their average fee per account compresses, the structural-threat thesis is confirmed. If flows stay positive and fees hold steady through the end of 2026, the AI disruption narrative was a sentiment event, and the stocks should recover their multiple as the curve re-steepens.
The IPO Freeze: A Canary for the AI Trade Itself
A third thread ran through the session, and it connects the AI story back to the bond market. SB Energy, the SoftBank-backed data-center developer that had been preparing one of the year's largest listings with a valuation north of $50 billion, postponed its IPO roadshow, according to people familiar with the marketing effort. The company had originally targeted a launch this month. Last week, nuclear services firm Holtec suspended its planned U.S. initial public offering indefinitely, citing what it called "an unusual confluence of developments that has impaired investor confidence in the market for new public offerings."
Both companies are AI infrastructure plays. Holtec's business is nuclear power for data centers; SB Energy's is the data centers themselves. When the two most prominent AI-capital-expenditure IPOs of the season stall within days of each other, it is not just a financing problem for two issuers — it is evidence that the marginal buyer of the AI trade is becoming selective about valuation.
This matters for financials through the capital-markets channel. A hot IPO market is fee income for investment banks, underwriting revenue for brokers, and equity-capital gains for the balance sheets that hold stakes in private deals. A frozen IPO market reverses all three. Raymond James and Schwab both run meaningful capital-markets and wealth-management platforms that benefit when private companies go public and founders diversify into managed accounts. If the AI capex boom pauses, the fee pipeline that financials have been underwriting their valuations on pauses with it.
The second-order implication is the one the market has not fully priced. If AI infrastructure companies cannot raise public equity at stretched valuations, they must turn to debt or to their parent companies' balance sheets. That pushes more financing demand onto the credit markets at a moment when the Fed is hiking and the curve is flat — precisely the conditions under which leverage becomes expensive and risky. The AI boom was funded on the assumption of endless cheap capital. The flattening curve and the IPO freeze together suggest that assumption is being tested.
What Comes Next: Scenarios Across Time Horizons
Short term (weeks): Sentiment is fragile and the technical setup is crowded. The next catalysts are the monthly employment report and the next inflation print, because both will move the front end of the curve and, with it, the net-interest-margin math for banks. If inflation prints hot, the front end rises further, the curve flattens more, and financials retest their lows. If inflation cools, the curve re-steepens and the oversold brokers rebound. This is a trading range driven by data, not a trend.
Medium term (quarters): The base case is that the Fed delivers one more 25-basis-point hike before pausing, as its own dot plot suggests, and that the curve stabilizes in the 15-to-30-basis-point range. In that scenario, bank earnings hold up because loan growth and credit quality remain solid — Sykes' point that "the fundamentals are good" and "there's a good economy and good employment." The exposed names are the pure fee-takers with high advisory concentration; the beneficiaries are the diversified money-center banks with strong deposit franchises and lower reliance on wealth-management fees.
Long term (years): This is where the structural call lives. If AI agents become the default interface for routine financial decisions, the advisory fee model compresses permanently, and the winners are the platforms that own the customer relationship and the data — which may or may not be today's brokers. If, instead, regulation, fiduciary complexity, and client inertia protect the incumbents, Tuesday's selloff will look like a buying opportunity in hindsight. The resolution of that question will take years, and no single data point settles it.
The upside case for financials is that the Fed hikes once more, inflation falls back toward target, the curve re-steepens to its historical norm, and AI agents turn out to be lead generators for human advisors rather than replacements. In that world, Schwab at a 6% discount is cheap. The downside case is that the Fed overtightens, the curve inverts, credit losses rise, and AI agents begin to displace advisory fees at the margin — in which case the multiple compression has only just begun.
The signal to watch is the one that ties both threads together: the 2s10s spread. If it holds above 15 basis points through the fourth quarter while broker flows stay positive, the structural bears are wrong and this was a cyclical air pocket. If it breaks below 10 basis points and inverts, the market is pricing a slowdown, and the AI disruption narrative will find fertile ground in weaker earnings.
Tuesday's selloff was not a verdict on the financial sector's survival. It was the market discovering that two of its favorite assumptions — that AI only cuts costs, and that the Fed can hike without flattening the curve — cannot both be true at the same time. The curve is the faster-moving variable, and it will decide the next leg.
Explore more exclusive insights at nextfin.ai.
