Published by @dailyanalysts | Market close data as of 4:00 PM ET, after-hours data through ~6:15 PM ET
| Index | Close | Change | % Change |
|---|---|---|---|
| S&P 500 | 7,675.70 | -1.58 | -0.02% |
| Dow Jones Industrial Average | 53,463.88 | -113.52 | -0.21% |
| Nasdaq Composite | 26,130.20 | -21.10 | -0.08% |
| Russell 2000 | 3,006.56 | -3.46 | -0.11% |
Sectors: Best: Industrials (XLI) +1.09% — FedEx Freight and C.H. Robinson led on freight-demand optimism. Worst: Healthcare (XLV) -1.00%, with Real Estate (XLRE -0.60%) and Consumer Discretionary (XLY -0.67%) also lagging as rate-sensitive groups digested the hotter PCE print. Tech (XLK +0.61%) and Energy (XLE +0.60%) held up into Nvidia's print.
VIX: The Cboe Volatility Index sat around 15.2-15.7 — near the low end of its range and actually down slightly on the day. In plain terms: options traders were not bracing for fireworks despite the biggest single-day earnings load of the quarter (Nvidia, Salesforce, CrowdStrike, Okta). That's a complacency signal — when the "fear gauge" doesn't move on a night like this, it usually means good news is already priced in, which raises the bar for follow-through buying.
Treasury yields: The 10-year yield ticked up to ~4.65-4.66% (+1-3bp) and the 30-year sits near 5.2%, still close to its highest sustained levels since before the 2008 financial crisis. The move came because the annual PCE inflation figure ran hot (headline +3.7% y/y vs. 3.6% expected) even though the monthly core reading was in line at 0.2%. Translation for the average investor: mortgage and auto-loan rates aren't coming down soon, and the Fed has no clean argument yet to cut — but the December rate-hike odds (73% per CME FedWatch) matter more than the ones for September (down to ~36-38%).
Key technical level: The S&P 500 is sitting about 1.8% below its own all-time high of 7,816.70, set earlier this month. 7,600 is the line that matters — hold above it and this is normal digestion after a strong run; a decisive close below it would signal the "AI-earnings rally" is losing steam, not just pausing.
The catalyst: Two data points collided today. First, the Fed's preferred inflation gauge (core PCE, released this morning) held at 3.3% y/y — in line, but still 1.3 points above target, keeping the Fed's September meeting a genuine coin-flip. Second, and far more important, Nvidia's fiscal Q2 report landed after the close: $96.22B revenue (vs. $92.17B expected), $2.22 adjusted EPS (vs. $2.10 expected), data-center revenue of $89B (+117% y/y), and — the number that mattered most — guidance for fiscal 2028 revenue growth of 70%, more than 25 points above the 44% Wall Street was modeling.
Narrative strengthened: "AI capex is broadening, not concentrating." Nvidia's non-hyperscaler business (AI clouds, industrial, enterprise customers) grew 138% y/y to $40.3B — faster than its hyperscaler revenue. Amazon separately committed to buy 2 million Nvidia GPUs. CFO Colette Kress said the top five hyperscalers' combined capex is guided to $1.3 trillion in 2027, up from $800 billion this year. This directly rebuts the bear case that the AI trade rests on four or five balance sheets.
Narrative challenged: The "SaaSpocalypse" — the idea that generative AI guts legacy software vendors — took a direct hit tonight. Salesforce posted Agentforce annualized revenue of $1.5B, up 240% y/y, and raised full-year guidance; CrowdStrike put up its best quarter in company history explicitly because AI threats are driving cybersecurity demand. Benioff's words: "This is not the SaaSpocalypse." Two independent data points from two different software categories is a real signal, not noise.
What most investors are overlooking: Nvidia's own gross margin is now under pressure from the very AI buildout it's fueling. Margin guidance stepped down to 74% next quarter and is guided to bottom at 71-72% in fiscal Q4 — specifically because of a memory (HBM/DRAM) shortage. CFO Kress said it plainly: "Memory scarcity today is being driven in large part by the AI buildout itself." Nvidia's supply commitments — mostly memory procurement — more than doubled quarter over quarter, from $119B to $279B. On top of that, $7.8B of this quarter's $53.95B net income was a mark-to-market equity gain on Nvidia's stakes in Intel and SpaceX (following a $15.9B gain in Q1) — a meaningful and growing share of "earnings" that has nothing to do with chip sales. The AI capex boom is starting to inflate its own input costs, and that shows up first in memory pricing (Trendforce projects DRAM contract prices up 260% in 2026), not in Nvidia's headline revenue line.
Real-world read-through: This ties directly to the bond market story dominating the back half of August. Treasury Secretary Bessent has doubled long-bond buybacks and reportedly may tap the Treasury's ~$935-950B General Account to suppress yields further — a move Stanley Druckenmiller, Bessent's own investing mentor, called "a subsidy to procrastination" in a Wall Street Journal op-ed, warning that "governments defending prices against fundamentals always lose." If that fight is lost and long yields push back through 4.73% (this month's high), it raises financing costs for the exact AI data-center SPV structures underwriting names like Nvidia, CoreWeave, and the neoclouds — the transmission channel this desk has flagged as the defining risk of this cycle.
Abercrombie & Fitch (ANF) +35.7% to $147.75 (from $108.90) was the single biggest percentage mover among recognizable names — bigger than Nvidia, Salesforce, or CrowdStrike. But look under the hood: $1.75 of the $4.17 EPS beat came from one-time IEEPA tariff refunds, and comparable sales were actually flat, below the +0.5% estimate, with Hollister down 3%. The broader trend this signals: a growing share of today's "beats" — Abercrombie's tariff refund, Meta's litigation settlement, J.M. Smucker's guide raise — are being driven by one-time legal and policy line items rather than organic demand. Investors extrapolating today's headline numbers into a durable retail or consumer turn should look through the noise first.
Fade an NVDA opening gap above $225 back toward the $210-215 pivot. Nvidia beat on every metric and popped roughly 5% after hours toward ~$220, but the stock has closed lower the day after earnings in each of its last four reports despite beating every time (per Bespoke Data) — a well-documented "sell the beat" pattern. Entry: fade $225-230. Target: $212. Invalidation: sustained close above $232. Timeframe: 1-3 days.
Use any post-earnings pullback in NVDA toward $195-205 to add. Three independent signals confirm the AI infrastructure cycle is accelerating, not slowing: (1) FY28 revenue growth guided to 70% vs. 44% consensus, (2) hyperscaler capex guided to $1.3T in 2027 from $800B this year, (3) Amazon's fresh commitment to buy 2 million Nvidia GPUs. Entry: $195-205. Target: $250 (through the May 2026 record of $236.54). Invalidation: weekly close below $185. Timeframe: 1-3 months.
Buy the "sell-the-pop" dip in Salesforce and CrowdStrike if Thursday's open fades their after-hours gains. Software has been left for dead all year on SaaSpocalypse fears — Salesforce is still down 22% year to date even after tonight's beat-and-raise, and options positioning (CNBC's Options Action framed the sector as "software earnings disappoint" even as both names beat) shows real skepticism remains. That gap between fundamentals and sentiment is the contrarian setup. Entry: CRM $210-220 / CRWD $195-200 on a fade. Target: CRM $245 / CRWD $225. Invalidation: CRM close below $195 / CRWD close below $180. Timeframe: 1-2 weeks.
Reiterate the standing GLD hedge. Gold just posted its best month since 1999 near $4,650/oz while bitcoin simultaneously rallied past $78,500 — two independent assets confirming the same debasement/duration-hedge trade at the same time the Treasury is actively fighting the bond market (Bessent's expanded buybacks, a floated $935-950B General Account intervention) and Druckenmiller warns that "governments defending prices against fundamentals always lose." Today's setup reinforces it further: yields rose, not fell, on a hot annual PCE print — exactly the environment gold hedges best. Entry: $415-430 (already active from a prior session). Target: $460-470. Invalidation: 10-year yield sustained below 4.30% for two weeks, or GLD below $400. Timeframe: 3-6 months.
Nvidia's own numbers just quietly confirmed the AI trade has entered an inflationary phase, not a disinflationary one — and almost nobody is framing it that way. Supply commitments (mostly memory) more than doubled quarter over quarter to $279B; gross margin guidance has now been cut for two straight quarters specifically because of an HBM/DRAM shortage Nvidia's own buildout is causing; and $7.8B of this quarter's $53.95B net income was a mark-to-market equity gain on Intel and SpaceX stakes, not chip sales. The market is celebrating "70% FY28 revenue growth guidance" as unambiguously bullish. The honest read is that Nvidia is increasingly monetizing financial engineering (equity stakes, vendor financing, credit backstops) and geological scarcity (memory) alongside chip sales. That means the highest-conviction read-through trade for the next two quarters isn't more Nvidia upside — it's memory pricing power (Micron, SanDisk, and the broader DRAM/HBM supply chain) and the credit spreads on the SPVs financing the AI buildout, not the chip itself. If DRAM prices really do run up 260% in 2026 as Trendforce projects, the AI capex story starts eating its own tail through margin compression well before demand ever slows — and in a world celebrating a 5%-after-hours pop, nobody is pricing that risk yet.
Sources: Nvidia Q2 FY2027 earnings live blog — CNBC · Salesforce Q2 FY2027 results — CNBC · CrowdStrike Q2 FY2027 results — CNBC · Meta settlement — CNBC · July PCE report — CNBC · Druckenmiller on Treasury buybacks — CNBC · Markets News, Aug. 26, 2026 — Investopedia · Schwab Market Update · Daily Market Snapshot — Edward Jones · US Stock Market Index — Trading Economics
This content is for informational purposes only and is not investment advice. All price levels, targets, and invalidation points reflect the analyst's judgment based on data available at time of publication and are subject to change.