The consensus arrived at the weekend with one explanation and considerable confidence in it. CNBC counted five forces behind a sell-off that left the S&P 500 and the Dow Jones Industrial Average closed in correction territory and stripped the Nasdaq Composite of its entire 2026 advance; the weekly summaries put the broad index more than 10% below its October high; and the interpretive frame aimed squarely at retirement savers is that index concentration in artificial-intelligence beneficiaries has turned the market into a leveraged position on compute spending. Under that reading the damage is sectoral, it is contained, and it ends whenever semiconductor multiples find a floor.
The behavior of the largest American bank argues against containment. JPMorgan Chase (JPM) closed Friday at $316.58, down 1.14%, finishing in the lower third of its session range on a sixth consecutive red close and at the very bottom of its 90-day range. Over five sessions the stock has given up 8.2% — a larger decline than Microsoft (MSFT) or Alphabet (GOOGL) sustained across the same stretch — during a week universally described as an unwind of artificial-intelligence exposure. Livemint reported that Nifty Bank rose roughly 16% month-to-date to a record intraday high of 64,485 on November 20, which makes it difficult to argue that banking assets as a category were under global pressure. Something specific to the American cost of money is being repriced, and it is not being repriced through semiconductors.
Where We Are in the Cycle
De-ratings of this magnitude tend to proceed in phases, and the phase boundaries are visible in the 30-session performance table rather than in any single close. The first phase takes down the assets whose valuations depend most heavily on capital committed today against cash flows promised years out: NVIDIA (NVDA) has fallen 40.4% over thirty sessions, Advanced Micro Devices (AMD) 27.9%, Meta Platforms (META) 20.1% and Alphabet 20.8%. The second phase is subtler and far more consequential for allocation, because it moves the de-rating into businesses whose earnings are a function of the price and availability of credit. Only two of the ten large-capitalization names in this group are higher over thirty sessions — Visa (V) at plus 6.6% and Apple (AAPL) at plus 4.0% — and both of them convert revenue into cash without asking the capital markets for permission.
| Symbol | Close | Change | Day range | 52-week range |
|---|---|---|---|---|
| AAPL | $347.96 | -1.25 (-0.36%) | $346.95–$354.29 | $243.42–$360.83 |
| MSFT | $493.08 | -5.22 (-1.05%) | $491.02–$503.75 | $349.20–$559.15 |
| NVDA | $137.18 | -8.81 (-6.03%) | $135.63–$143.73 | $135.63–$243.37 |
| GOOGL | $278.01 | +5.35 (+1.96%) | $272.88–$283.06 | $272.11–$408.61 |
| AMZN | $253.78 | -3.59 (-1.39%) | $251.16–$259.95 | $196.00–$287.20 |
| META | $575.76 | -20.46 (-3.43%) | $571.64–$588.58 | $520.26–$779.82 |
| TSLA | $360.31 | -14.10 (-3.77%) | $359.99–$373.88 | $297.38–$498.83 |
| JPM | $316.58 | -3.64 (-1.14%) | $315.90–$320.79 | $279.10–$366.50 |
| V | $408.32 | +6.07 (+1.51%) | $402.42–$408.71 | $293.89–$442.23 |
| AMD | $441.72 | -47.47 (-9.70%) | $437.66–$474.72 | $188.22–$658.52 |
A month ago I argued that NVIDIA's slide was mechanical de-grossing that would reverse above $186.64. That call expired wrong, and the reason it was wrong is the organizing fact of this correction: what looked like a flow event was the opening leg of a capex-cycle re-rating, in which the market stopped paying forward multiples for infrastructure commitments it can no longer underwrite. Admitting that error has a direct consequence for how the second phase should be read. The impulse now will be to treat JPMorgan's six-session decline as spillover sentiment that mean-reverts within a fortnight; the earlier mistake suggests treating it instead as information about the cycle.
What the Session Revealed
AMD supplied the session's most instructive case. The stock gapped 3.52% lower, never traded above $474.72 at any point, and closed at $441.72 near the bottom of its range for a 9.70% loss — yet even after the worst single-session decline in the group it remains 4.1% above its 90-day low. NVIDIA, by contrast, closed at $137.18 after printing a fresh 52-week low at $135.63, leaving it 43.3% below its 90-day high and sitting precisely on the floor of that range. The faster-falling name has more room beneath it; the slower-falling name has already found the basement. That ordering is what a repricing looks like when the market is discriminating between degrees of capital intensity rather than selling a theme indiscriminately.
Over five sessions JPMorgan underperformed Microsoft by 270 basis points — a commercial bank losing more ground than the largest software franchise in a week attributed entirely to artificial intelligence.
The two advancers reinforce the same distinction without contradicting it. Alphabet gained 1.96% to $278.01 and closed mid-range, which after a 20.8% thirty-session decline is a bounce inside a de-rating rather than evidence the de-rating has concluded. Visa added 1.51% to $408.32 and finished essentially at its session high, having held a 6.6% gain across a period in which the index entered correction. Meanwhile Livemint reported that Taiwan widened curbs on short sales and has been purchasing shares after the island's benchmark suffered its worst weekly fall in 35 years. Administrative intervention of that kind generally appears when a repricing has already done most of its work on the complex in question, which argues that semiconductors are late in their leg and financials are early in theirs.
The Positioning Implication
The obvious defensive answer inside mega-cap equity is Visa and Apple, and the obvious answer is already in the price: Visa trades 14.0% above its 90-day low and only 5.6% below the high, while Apple sits 2.8% below its 90-day high and within 3.6% of its 52-week high of $360.83. Paying up for demonstrated resilience after the resilience has been demonstrated is how investors convert a good idea into a mediocre risk-adjusted return. The genuinely mispriced asset in this group is Microsoft, which closed at $493.08, 10.9% beneath its 90-day high and a mere 1.2% above the low, down 7.5% over thirty sessions. The market is currently valuing Microsoft as an infrastructure capital program when its cash conversion profile more closely resembles the two names investors are crowding into for safety, and that misclassification is the asymmetry worth owning. The $415.00 marker I set on Visa on Saturday remains live into December, though that argument was about which cohort held, not about where the next leg of damage lands.
The thesis lives or dies on JPMorgan. A close below $310.00 within ten sessions would confirm that the correction has migrated into balance-sheet-sensitive earnings and that financial-sector allocations built on a steady rate path require revision. Microsoft's reclaim of its Friday high at $503.75 within fifteen sessions would validate the view that quality cash generation separates from capex intensity as the de-rating matures; a close beneath its 90-day low of $487.32 over the same window would mean the repricing is indiscriminate after all and the distinction I am drawing does not exist. What would change my mind outright is simple to specify: JPMorgan recovering above its Friday high of $320.79 and holding there while semiconductors continue to make new lows. That sequence would prove this is a contained sectoral event and that the bank's six-session slide was noise.
The Weight of Evidence
The market is pricing an artificial-intelligence accident; what is actually underway is a broad repricing of the discount rate applied to every long-duration earnings stream, which began in compute and has now reached credit. Investors treating this as a semiconductor problem will be positioned for the wrong second act, because the names that have already fallen 40% carry considerably less unrecognized risk than a bank at the bottom of its ninety-day range on a six-session losing streak. The allocation that makes sense here concentrates in cash-generative franchises that have been incorrectly bundled with the capex complex — Microsoft foremost among them — while treating financial-sector weakness as the leading indicator it appears to be. I expect JPMorgan to make new lows before NVIDIA makes new highs.