Everything About AI
NextFin News - Shares of America's largest banks are on track for their worst weekly losing streak since the 2008 financial crisis, as fears over artificial-intelligence lending, stubbornly high Treasury yields, and sticky inflation battered the sector through the week. The headline, however, masks a sharper truth: the damage was concentrated, not broad. Bank of America, Morgan Stanley and Citigroup each lost more than 4% over the five sessions, while JPMorgan Chase and Wells Fargo barely moved - a split that points to exactly where the market's fear is living, and where it is not.
The week that closed Friday capped a brutal stretch for financial shares. The Invesco KBW Bank ETF, the most direct proxy for the big-lender complex, ended the week lower, and the broader bank-stock gauge logged its weakest weekly performance since the depths of the global financial crisis. The selloff arrived despite a solid start to third-quarter earnings season, with JPMorgan Chase, Goldman Sachs, Wells Fargo and Citigroup all reporting higher profits this week - a reminder that in today's market, good news is not enough when the macro backdrop turns hostile.
The numbers tell the story of a sector at war with itself. Bank of America closed the week at $51.52, down 5.2% from the prior Friday; Morgan Stanley finished at $181.04, down 4.7%; Citigroup ended at $123.95, down 4.4%; Goldman Sachs closed at $881.61, down 1.5%. By contrast, JPMorgan rose 1.7% to $338.62 and Wells Fargo finished essentially flat at $83.68. The financial sector as a whole, measured by the Financial Select Sector SPDR, actually gained 1.1% for the week, and the regional-bank gauge added 1.2% - evidence that the selling was selective rather than indiscriminate.
The AI Financing Chain Is Where the Fear Lives
The immediate trigger sits squarely in the artificial-intelligence funding boom that has powered both Wall Street profits and, increasingly, Wall Street's anxieties. On Wednesday, shares of Oracle fell more than 5% after reports that Blue Owl Capital, the private lender that has been Oracle's biggest data-center financing partner, would not back a $10 billion deal for its next facility. The episode followed an earlier warning shot: on September 24, Oracle invoked a force majeure clause tied to Blue Owl's Stargate data-center campus in New Mexico, part of a $165 billion buildout across Texas, New Mexico and Ohio. Blue Owl itself has lost more than 45% of its value over the past year, and private-credit defaults have climbed to a reported record 6.3%.
Oracle's move to protect itself against potential delays at a massive data center project sent a chill through the trillion-dollar market for AI infrastructure financing, according to interviews with bankers and investors.
For banks, the transmission mechanism is direct, and it runs through three channels. First, the origination channel: big banks have underwritten and arranged billions of dollars of debt for data-center developers, hyperscalers and the utilities that power them. When a marquee project stalls, the pipeline of future fee income shrinks. Second, the credit channel: loans already on the books were underwritten against revenue projections that assumed insatiable demand for AI computing. If that demand disappoints, borrowers struggle to service debt, and provisions for credit losses rise. Third, the collateral channel: data centers and AI infrastructure are highly specialized assets with a thin resale market. A loan secured by a half-empty data center is worth far less than the same loan secured by an office tower in a major city - and far harder to sell at any price.
The market is not pricing a broad credit event; it is repricing the AI-exposed slice of bank balance sheets, and doing so abruptly. A loan book that looks pristine in a boom can deteriorate quickly when the revenue assumptions behind it fail to materialize, and the speed of the repricing reflects how much new AI-related lending banks have added since the technology boom began.
The concern is amplified by the sector's own leadership. JPMorgan Chase chief executive Jamie Dimon has warned that AI capital spending could reach $1 trillion in 2027 - a sum he described as "like 1% increase to GDP each year" - while cautioning that the spending "may add a little bit to inflation." When the head of the largest US bank frames the AI boom as both an inflationary force and a historical echo of past technology bubbles, investors listen. Dimon has also cautioned that it remains premature to identify the winners of the AI race, citing the internet bubble as a template for how the industry could evolve - a pointed reminder that the companies building the infrastructure today are not guaranteed to be the ones capturing the profits tomorrow.
Yields Are the Second Blade
While AI anxiety supplied the spark, the bond market supplied the fuel. The 10-year Treasury yield, which drives the discount rate for bank earnings and the mark-to-market value of bank bond portfolios, traded at 5.105% on Friday after touching 5.31% earlier in the month - its highest level since April 2002. Yields rose on a hotter-than-expected inflation print and a resilient labor market that together dimmed expectations for near-term Federal Reserve easing. For a sector whose valuation is unusually sensitive to the level of long-term rates, a 10-year yield above 5% is not a headwind; it is a re-rating event.
High yields cut banks two ways. First, they compress net interest margins: the gap between what banks pay depositors and what they earn on loans has narrowed as the 10-year-to-2-year Treasury spread sits near 0.46%, down from 0.74% in February. A flatter yield curve means banks borrow short and lend long for a thinner profit, and the compression is most acute for the consumer and commercial lenders that dominate the indexes that fell hardest this week. Second, they depress the market value of the securities banks hold, creating unrealized losses that linger on balance sheets long after the 2023 regional-banking crisis faded from headlines. Those losses do not hit earnings unless the assets are sold, but they constrain capital flexibility and remind investors how quickly a rate shock can turn a healthy-looking balance sheet fragile.
There is also a competitive dimension that runs through the deposit base. Economist Torsten Slok has argued that the risk runs deeper than the current cycle: AI agents could soon automate the movement of household cash from checking accounts paying 0.1% into accounts offering 3.3% to 5.0%, forcing banks to compete harder for deposits just as their lending margins thin. Cheap, sticky deposits are the foundation of bank profitability; if technology makes those deposits footloose, the entire economics of retail banking shifts.
The regulatory shadow reinforces the point. On October 1, Federal Deposit Insurance Corporation chairman Travis Hill said the agency would not rule out the risk that AI tools could allow depositors to move funds faster in search of higher returns - a modern, automated version of the bank run - and made what he called a "strong case" for expanding deposit insurance coverage. A regulator publicly flagging AI-enabled deposit flight is not a scenario banks want priced into their shares, because the response - higher insurance assessments, stricter liquidity rules - would come directly out of future profits.
Cyclical Selloff, Structural Question Mark
Is this the start of a structural bear market for banks, or a cyclical air pocket? The evidence points to cyclical - with a structural asterisk that investors cannot afford to ignore.
The cyclical case rests on three pillars, and each has historical support. First, the selloff is concentrated rather than systemic. The laggards - Bank of America (-5.2%), Morgan Stanley (-4.7%), Citigroup (-4.4%) - are the names most exposed to wealth management, capital markets, and consumer lending, the businesses most sensitive to a risk-off turn. The leaders - JPMorgan (+1.7%) and Wells Fargo (flat) - are the diversified, deposit-rich franchises that can ride out a volatility spike. A systemic crisis does not discriminate; this one did.
Second, the trigger is a positioning shock, not a solvency shock. No major bank has reported deteriorating credit quality this week; the decline follows a valuation reset in AI-linked assets, not a wave of defaults. Third, history offers a template for mean reversion. In February 2026, private-credit worries pushed financials to their worst day since April's tariff turmoil, and the selloff reversed within weeks as earnings proved resilient. In March 2023, regional banks fell on deposit-flight fears, and the survivors recovered once liquidity backstops calmed the system. In 2018, bank stocks sold off hard as yields rose and the yield curve flattened, only to rally strongly in 2019 once the Federal Reserve paused. Each episode looked existential in the moment; each proved cyclical in retrospect.
The structural asterisk matters, though, because it attacks the foundation rather than the price. If AI-driven deposit disintermediation becomes real - if agents do begin moving household balances at scale - the low-cost deposit franchise that underpins bank profitability would face a permanent repricing. That is not a cyclical headwind that mean-reverts; it is a regime change that resets the baseline. The same logic applies to AI lending: if the current buildout proves to have been a capital misallocation on the scale of the dot-com fiber-optic buildout, the loans behind it will not "recover" - they will default. But neither outcome is yet observable in the data, and the banks best positioned - those with deep deposit relationships and diversified revenue - are precisely the ones the market left untouched this week.
The Counter-Thesis: The Selloff Is Overdone
The strongest argument against the bearish read is simple: the earnings are fine. Third-quarter Finance sector earnings are expected to rise 7.2% from a year earlier on 7.5% higher revenue, with the four biggest banks all reporting higher profits this week. JPMorgan, Goldman Sachs, Wells Fargo and Citigroup each posted year-over-year profit gains, and the growth momentum is broad-based - 14 of the 16 major industry sectors are expected to deliver positive earnings growth this quarter. By that measure, the sector is growing into its valuation, not away from it. A selloff on solid earnings is, in the classic framing, a buying opportunity, and investors who sold this week may be giving away cheap exposure to a profitable industry.
That argument underestimates how much of bank valuation rests on what comes next, not what just happened. Equity investors are forward-looking discounting machines, and this week they discounted three things at once: the possibility that AI lending turns sour, the certainty that yields stay higher for longer, and the risk that regulators respond to AI-driven deposit mobility with costlier insurance requirements. Solid backward-looking earnings do not offset forward-looking fear - they merely establish the baseline from which the de-rating occurs. The counter-thesis is right that no credit crisis has arrived; it is wrong to conclude that the market's repricing is therefore irrational. Prices reflect the future, and the future, this week, looked more expensive and more uncertain than it did a week ago.
The falsifying signal is specific, and it cuts both ways. If the 10-year Treasury yield falls back below 4.75% and the KBW bank index recovers its weekly losses within two trading weeks, the cyclical read is confirmed and the structural fears were noise. Conversely, if Bank of America, Citigroup and Morgan Stanley each fall another 10% on explicit warnings about AI-related credit or deposit outflows, the structural read strengthens and this week becomes the first chapter of something longer. Watch the next two quarters of loan-loss provisions and net charge-offs: they will tell investors whether the AI financing boom is turning toxic or merely maturing.
What Comes Next
In the short term, expect volatility to persist. Bank shares are caught between an earnings season that is delivering and a macro backdrop that is not cooperating; until the 10-year yield finds a ceiling, rallies are likely to be sold. The medium-term path depends on credit quality: the next two quarters of loan-loss provisions and net charge-offs will tell investors whether the AI financing boom is turning toxic or merely maturing. Over the long term, the structural question - whether AI reshapes deposit behavior and lending demand - will dominate, and the winners will be the lenders with the strongest deposit franchises and the least exposure to speculative technology credit.
Three scenarios frame the road ahead. In the base case, yields stabilize near current levels, AI lending losses remain contained, and bank shares grind sideways as earnings growth offsets multiple compression - a stock-picker's market where the divergence between the winners and losers widens. In the upside case, inflation cools, the Federal Reserve signals a clearer easing path, and the AI credit scare proves to be a single-project issue rather than a sector-wide problem; bank shares could then reclaim their early-October highs, led by the same names that held up this week. In the downside case, private-credit defaults accelerate beyond the current 6.3% level, AI borrowers begin to miss payments, and deposit outflows materialize; the KBW index would then test its 2026 lows, and the cyclical label would no longer apply.
The week's lesson is not that banks are broken. It is that the market has stopped rewarding them for the AI boom they helped finance and started charging them for the risk they took on. Bank of America, Morgan Stanley and Citigroup are paying that price today; JPMorgan and Wells Fargo, for now, are not. The divergence is the story - and it will narrow only when yields stop rising or credit starts to crack.
Explore more exclusive insights at nextfin.ai.
