Daily U.S. Market Review — Friday, June 26, 2026 (after the close) · @dailyanalysts
Friday looked boring — the S&P 500 closed down a rounding-error 0.05% — but that calm masked the most important rotation of 2026: money is fleeing AI infrastructure and piling into healthcare, which just had its best week since 2022. The week's real damage was brutal: the Nasdaq fell 4.6%, its worst week in a year, and Oracle logged its worst week since the 2001 dot-com bust. My stance: cautiously constructive on the broad market, decisively bearish on leveraged AI-capex names, and bullish on the healthcare/defensive leadership baton-pass — call it 7/10 conviction. This is a rotation, not yet a top — but the leadership that carried the market for two years is now the problem.
| Index | Close | Day | Week |
|---|---|---|---|
| S&P 500 | 7,354.02 | -0.05% | -2.0% |
| Nasdaq Composite | 25,297.62 | -0.24% (5th straight loss) | -4.6% |
| Dow Jones | 51,876.11 | -0.09% | +0.6% |
| Russell 2000 | — | +0.3% (only green index) | positive |
Best sector: Health care +3.0% on the day (+7%+ on the week — best week since June 2022). Staples, real estate, discretionary, utilities and financials all closed green.
Worst sector: Technology -1.9%, with industrials -1.6% close behind.
VIX: ~20.1, up roughly 6.5% intraday. In plain terms: the "fear gauge" climbed off its complacent spring lows but is nowhere near panic (panic is 30+). Translation — investors are nervous and hedging, but not running for the exits. This is anxiety, not capitulation.
Treasury yields: The 10-year eased to about 4.37% and the 2-year fell ~3bp to 4.09%. Falling yields as inflation runs hot (core PCE 3.4%, a 32-month high) is unusual — it's the bond market betting new Fed Chair Kevin Warsh's tough-on-inflation talk will work. The catch: the 2s/10s curve is flattening fast. If it inverts, recession alarms start ringing.
The catalyst: A New York Times report that OpenAI may delay its IPO to 2027 — citing SpaceX's sloppy post-IPO performance (back near its $150 debut) and AI-stock volatility — lit the fuse. Wall Street instantly questioned whether the capital markets will keep funding the AI build-out, hammering the companies spending the most: chips and Oracle. Asia was worse — Korea's Kospi fell 5.8% and tripped circuit breakers; SoftBank plunged 12%+.
The narrative that got challenged: "AI capex is bulletproof — buy every dip in semis." This week broke it. Citi cut technology to market-weight, with strategist Scott Chronert warning it's "difficult to see how everyone in the AI/Tech path wins." Jeremy Grantham called this the most expensive U.S. market "in American history." Even Micron — which printed record revenue and EPS Wednesday — couldn't hold its gains, falling nearly 7% Friday.
The narrative that got stronger: Broadening is healthy. Advancing stocks outnumbered decliners even on down days; 63% of the S&P now trades above its 50-day (up from 50% at the start of June), and the equal-weight S&P is beating the cap-weighted index. A market that no longer needs seven stocks to rise is a sturdier market.
What most investors are overlooking: This isn't really about an IPO calendar. Analysts at Jefferies, Morgan Stanley and Deutsche Bank flagged a "second DeepSeek moment": Z.ai's new GLM5.2 model is "almost equal to Anthropic… at one quarter of the cost per token," while DeepSeek's V4-Pro reportedly does ~90% of everyday tasks at ~1.5% of the cost of Anthropic's flagship. The threat isn't AI demand — it's AI pricing power. If the price of intelligence is collapsing, the ROI math under $300B+ of hyperscaler-and-Oracle capex gets a lot worse.
Real-world link: Memory-chip scarcity is already leaking into your wallet — Apple and Microsoft both raised hardware prices this week citing chip costs. Those costs could feed core-goods inflation into Q3–Q4 just as the Iran energy spike fades, giving Warsh cover to hike even as oil sits below $70.
| Stock | Close | Day | Why |
|---|---|---|---|
| Eli Lilly (LLY) | $1,208.12 | +7.1% | Healthcare rotation's flagship; GLP-1 franchise leadership. Powered the sector's best week since 2022. |
| Microsoft (MSFT) | $372.97 | +5.7% | Oversold bounce after its worst June since 2000 (-25% YTD). Dead-cat or bottom? Either way, a violent reversal. |
| Palantir (PLTR) | $112.93 | +5.3% | Tagged "most oversold mega-cap tech" — a mechanical snap-back, not new news. |
| J&J (JNJ) / UnitedHealth (UNH) | $254.66 / $427.89 | +4.0% / +3.0% | Defensive healthcare bid; UNH pressing a 52-week high on a quiet turnaround. |
| Stock | Close | Day | Why |
|---|---|---|---|
| Micron (MU) | $1,132.33 | -6.7% | Gave back its post-earnings pop despite record results — punished by the AI-capex scare, not by fundamentals. |
| Broadcom (AVGO) | $365.02 | -3.7% | Caught in the broad semi rout; the AI-networking bull case got overwhelmed by sector risk-off. |
| Oracle (ORCL) | $148.53 | -2.6% | Worst week since 2001 (-19%), now -55% from its September peak. $130B debt, -$24B free cash flow, capex +162%. The poster child for "spend-to-win" anxiety. |
| Nvidia (NVDA) | $192.53 | -1.6% | Worst week in over a year (-8%) as money rotated from chips into defensives. |
Most surprising mover — ON Semiconductor (-14%): ON cratered on the day it announced its largest-ever acquisition, a $7B all-stock buy of Synaptics. A year ago that's a "visionary expansion" pop. Today the market punished it. The signal: investors have flipped from rewarding growth-by-spending to demanding discipline and returns. That same instinct is what's crushing Oracle and pressuring the hyperscalers — the entire "capex is good" reflex of the AI era is being repriced in real time.
① ACTIONABLE TOMORROW — Stay long / add Healthcare (XLV). LONG-TERM HIGH CONVICTION
Two independent signals agree: the best week since 2022 and improving market breadth. This is real leadership rotation, not a one-day pop. Entry on any pullback $156–160 (closed $160.34), target $172, invalidation daily close <$150. Names: LLY, UNH, JNJ.
Scorecard: I flagged "XLV add $153–156, target $165" on June 25. It's now $160.34, +3% Friday — working.
② CONTRARIAN — Accumulate Micron (MU) into the capex panic. SHORT-TERM / SWING (1–3 mo) SPECULATIVE
The crowd is selling the best fundamental story in tech — record revenue, record EPS, a genuine memory shortage — on a sentiment scare about other companies' spending. Memory is the one part of the AI stack with physical scarcity and pricing power regardless of which model wins. Entry $1,050–1,135 (closed $1,132), target $1,400, invalidation daily close <$980. Higher risk; size accordingly.
This is a 1-signal contrarian bet against momentum — keep it speculative.
③ DEFENSIVE — Build a staples/utilities sleeve and buy a little protection. LONG-TERM WATCH/PROTECT
BofA's technical desk is urging clients to add protection into Q3. Establish a defensive sleeve in consumer staples (XLP, $84.71) and utilities (XLU, $46.20) — both green Friday — and set a mechanical rule: trim hyperscaler/AI-capex exposure if the S&P closes two days below 7,356. Note: avoid reaching for long-duration Treasuries as your hedge here — a Warsh hike is now live (Kashkari said Friday he expects one hike this year), and that's the wrong risk to add.
Most important event: June Nonfarm Payrolls, Thursday July 2, 8:30am ET (pulled forward ahead of the July 3 holiday close), with ISM Manufacturing July 1 and JOLTS/Consumer Confidence June 30. Early forecasts see June job growth slowing sharply from May's 172,000. With the market now "preoccupied with inflation," a hot wage print could pull a Fed hike forward to September; a soft one buys the rotation more room.
Key price level: S&P 500 7,356 (50-day MA). Hold it and the broadening thesis stays intact toward 7,800–8,100 bull targets. Lose it for two sessions and a Q3 correction becomes the base case.
Three to keep on the radar:
Wall Street is mispricing why tech sold off. The consensus story is "OpenAI delayed its IPO" and "AI capex got too expensive." The real, under-discussed driver is that the price of AI itself is collapsing. Z.ai's GLM5.2 matches Anthropic at a quarter of the cost per token; DeepSeek's V4-Pro reportedly handles 90% of tasks at ~1.5% of flagship pricing. That does not kill AI demand — it shreds the margin and pricing-power assumptions baked into hundreds of billions of infrastructure spend.
So the 2H 2026 trade isn't "bubble pop vs. buy the dip." It's a bifurcation: short the leveraged capex-payers whose ROI math just got worse (Oracle, with $130B debt and -$24B free cash flow, is the cleanest expression), own the physical bottleneck where scarcity still commands price regardless of which model wins (memory — Micron), and own the cash-flows that don't depend on token economics at all (healthcare). Cheap-model deflation is quietly bullish for AI adopters and the broadening rally, and bearish for indebted AI landlords. That split — not the index level — is where the money will be made and lost into year-end.
Sources: CNBC market close · CNBC: Oracle's worst week since 2001 · CNBC: "another DeepSeek moment" · CNBC: Citi cuts tech · Schwab Market Update (VIX, yields, 50-day) · CNBC: week ahead · CNBC: U.S. strikes Iran. Prices verified at Friday's close, June 26, 2026.
This is analysis and opinion for informational purposes, not personalized investment advice. Do your own due diligence.