Micron just detonated the AI-bubble thesis after the bell. A quadrupling of revenue, a Q4 guide of $50 billion (vs. ~$43.6B expected), 84.9% gross margins, and $22 billion of locked 3–5 year contracts — the loudest possible answer to two days of "AI is cracking" panic imported from Korea's KOSPI crash. The cash session itself was a quiet, healthy rotation: Dow and small-caps green, tech soft for a third straight day, oil collapsing to pre-war levels.
Stance: cautiously bullish on the AI-infrastructure complex, neutral on the broad tape — confidence moderate-to-high. The one thing that can spoil it lands in 16 hours: Thursday's core PCE.
| Index | Close (approx) | Change |
|---|---|---|
| S&P 500 | ~7,341 | -0.1% |
| Dow Jones | ~51,857 | +0.4% (+~190 pts) |
| Nasdaq Composite | ~25,490 | -0.4% |
| Russell 2000 (IWM) | $296.69 | +0.5% |
ETF proxies: SPY $733.24 (-0.05%), DIA $518.52 (+0.37%), QQQ $710.62 (-0.42%), IWM $296.69 (+0.46%).
Best sectors: Industrials +1.16%, Consumer Discretionary +1.15%, Utilities +1.04%, Staples +0.86%.
Worst sectors: Energy -1.63% (oil crash), Communication Services -0.68%, Technology -0.62%.
The tell: three down days for the Nasdaq, yet 7 of 11 sectors closed green and small-caps led. That is rotation, not liquidation. Schwab's breadth gauge actually improved — stocks above their 50-day MA rose to 58% from ~50% even as the mega-cap leaders fell.
The VIX backed off Tuesday's 19.5 spike, easing toward the 18–19 area. In plain terms: after Tuesday's scare, traders stopped paying up for crash insurance. Tech-specific volatility stayed elevated relative to the broad market — investors were hedging a tech shakeout, not a market one. That distinction matters: it means the panic was contained to a crowded trade.
The 10-year yield fell to ~4.41% from 4.50%, and WTI crude crashed ~5% to ~$69.80 — its lowest since the Iran war began, now within a few dollars of its pre-war price. Lower oil = lower future inflation = lower odds the Fed has to hike. But the 2-year sits at 4.16%, pinned near its 2026 high, because the market still prices at least one Warsh rate hike this year. Above 4.55% on the 10-year, AI multiples compress ~10–12%; back below 4.30% and the rally has clear air. Today's drop to 4.41% bought the bulls room — but PCE can erase it overnight.
The catalyst was a triple-header of AI silicon news, capped by Micron after the close.
Narrative strengthened: the memory/AI supercycle is real and supply-constrained through 2028. Monday–Tuesday's $900B chip wipeout (KOSPI -10%, a "Black Tuesday") was crowded-positioning unwind, not demand collapse. Micron's $22B of binding contracts is the receipt.
Narrative challenged: "AI capex is peaking / ROI doubts." Hyperscaler 2026 capex growth has gone from a forecast +40% to +75%+ year-over-year (per JPMorgan). The doubt isn't demand — it's who pays.
Real-world implication: the second-order winner of the AI boom is now electricity. Sunrun jumped 12.6% on a Tesla deal to power AI data centers with battery storage; utilities (+1.0%) were a top sector. AI's constraint is shifting from chips to power and memory — both inflationary, both supply-limited.
Most surprising mover & what it signals: Sunrun (RUN) +12.6% on a Tesla AI-data-center storage deal. A near-bankrupt-narrative solar name surging on AI demand tells you the AI trade is broadening into its power/energy-storage supply chain — the next leg most portfolios don't own. (Tesla itself fell 1.6% to $375.53.)
Micron is already +13% after hours; the easy money there is made. The cleaner entries are the memory peers that haven't repriced — SanDisk (SNDK) and Western Digital (WDC) — plus Nvidia (NVDA) on the confirmed demand. NVDA: accumulate $190–200 (closed $199), target $225, invalidation daily close <$182. Timeframe 1–2 weeks. For short-term traders. Use a hot-PCE gap-down as the gift, not a reason to abandon the thesis.
The crowd sees a late-to-data-center modem company; today it named Meta, set a $15B data-center target, raised 2029 EPS guidance to $18+ (vs. $15.26 consensus), and bought a CUDA alternative. The +15% after-hours move is the start of a re-rating, not the end. Accumulate QCOM $195–215 (it will gap up), target $250, invalidation daily close <$180 or a named hyperscaler deal slipping. Timeframe 1–3 months. For long-term investors.
With PCE a binary risk and a Warsh hike on the table, you want low-beta protection — but utilities give you a free option on the AI-power theme (see Sunrun). XLU add $45–46 (closed $45.54), target $48, invalidation close <$43.50. Timeframe 1–3 months. For long-term/defensive investors. It cushions a rate-shock day while the rest of the book stays long AI infrastructure.
Most important event: Thursday, June 25, 8:30am ET — May Core PCE (the Fed's preferred gauge), alongside Q1 GDP final (~1.6%) and jobless claims. Consensus: headline PCE +4.1% y/y, core +3.4% y/y / +0.3% m/m — which would be the hottest core since October 2023. Core at or above 3.4% hands Warsh his hike rationale; a soft print (≤3.2%) lets the 10-year stay near 4.41% and frees the rally. Treasury Secretary Bessent has already floated a single "tap the brakes" hike — read that as a White House green light.
Key level to monitor: S&P 500 7,140. A close below 7,090 invalidates the two-month bull flag and turns the diamond reversal live toward a deeper pullback. Above 7,140, dips are buyable toward the 7,680 flag target.
Three for the radar:
Tonight's headline everywhere will be "Micron blowout = AI bull confirmed." The deeper, less-discussed truth: Micron just told you the price of AI is going up, and it locked it in through 2028. The $22B of binding multi-year memory contracts and "shortage into 2028" language mean rising input costs for every hyperscaler and AI-software company — which is exactly why Microsoft (-2.3%) and Oracle (-4.6%) fell on the same day Micron soared. The trade is not "buy AI." It is own the bottlenecks with pricing power — memory (MU/SNDK/WDC), custom silicon (AVGO/QCOM), and power (utilities) — and fade the cost-takers (AI software, services, and margin-squeezed hyperscalers). And keep one hand on the exit: the swing factor in the next 24 hours isn't AI at all — it's whether core PCE prints hot enough to give Warsh his hike. A blowout from Micron will not save the multiple if the 10-year rips back through 4.55%.
NVDA accumulate $190–200 (called 6/23): WORKING — NVDA at $199 today, squarely in the zone, and Micron's print just validated the demand leg of the thesis. MU "don't chase into the print" WATCH (6/23): the risk call was right — it dodged Tuesday's -13% — but the fundamental was even stronger than feared; the lesson is to buy memory on rate-driven dips, not headline fear.
Primary sources: CNBC — Micron Q3 results · CNBC — Qualcomm Investor Day · CNBC — OpenAI/Broadcom Jalapeño · Charles Schwab Market Update · Investopedia Markets Today · CNBC/CappThesis — S&P chart patterns · CNBC — S&P flash PMI / factory job cuts.
This is analysis and opinion, not personalized investment advice. Do your own due diligence. Levels and targets reflect the author's judgment as of June 24, 2026.