The Invoice, Settled: Microsoft’s Azure Bought an 8% Rally, Meta’s 91% Cash-Flow Collapse Cost 8%

1. Macro Pulse

The setup: a relief rebound with a fault line running through it. Microsoft’s Azure acceleration rescued the AI-demand narrative and dragged semiconductors up more than 3% pre-market, but the 30-year Treasury at a two-decade high is a warning the equity bounce is choosing to ignore. J.P. Morgan’s positioning monitor flashed “oversold enough to warrant a tactical buying opportunity,” which is the bull case in one line. [CNBC/TheStreet, 07/30]

Wednesday’s close was ugly: the Dow fell 1,153.18 points (−2.19%) to 51,594.14, its worst day since April 2025. The S&P 500 slid 1.52% to 7,316.15 and the Nasdaq Composite fell 1.74% to 24,442.94, ending more than 10% off its all-time high, in a technical correction. The trigger was not the hold; it was Warsh’s refusal to signal any tolerance for above-target inflation. [CNBC, 07/30]

The Fed and the bond revolt: the FOMC held at 3.50–3.75% for a fifth straight meeting, but Warsh told reporters the Fed “will not hesitate” to act on inflation, that it “will deliver on price stability,” and pointedly that “there is no soft inflation target.” The bond market read that as a Fed potentially behind the curve: the 30-year yield jumped 11 bps on the decision and reached 5.23%, its highest since 2007, and the curve steepened sharply. CME FedWatch now prices about a 76% chance of a September hike. [CNBC/Bloomberg/Schwab, 07/30]

This morning’s data, released 8:30 AM: Q2 advance GDP rose just 1.5% annualized, down from 2.1% in Q1, as government spending fell and investment decelerated. The Q2 PCE price index rose 5.1% and the core PCE deflator 3.4%. The combination, slowing growth with sticky inflation, is the stagflation tension the Fed is now navigating without a dot plot. [BEA, 07/30]

Commodities: Brent trades near $90 after rallying almost 8% Wednesday on renewed US strikes on a dozen Iranian targets; WTI above $84. Gold consolidates near $4,020 to $4,050, capped by the surge in real yields even as growth data soften, the same real-yields-over-danger regime that has held all cycle. [Bloomberg/FXStreet, 07/30]

An Asian tell: Zhongji Innolight, a Chinese data-center optical-parts maker, fell 5% on its Hong Kong debut after raising $6.8B, Asia’s second-largest listing this year behind CXMT. Even a hot AI-infrastructure IPO is now getting a cool reception, a small sign the AI enthusiasm that drove issuance is being repriced. [CNBC, 07/30]

2. Economic Calendar

Also reporting today: Mastercard, Shell, Bristol-Myers Squibb, Altria and Stryker before the open; Apple and Amazon after the close. Tomorrow brings ExxonMobil and Chevron, then the July jobs report looms next week. [Schwab, 07/30]

[ANALYSIS] The GDP-and-PCE pairing is a genuine problem for the Fed and the market. Growth at 1.5% says the economy is cooling; core PCE at 3.4% says inflation is not. A Fed that hikes into 1.5% growth risks a policy error toward recession; one that holds into 3.4% core risks losing inflation credibility, which Warsh clearly refuses to do. That is why the long end sold off: the bond market thinks the Fed will prioritize inflation and accept slower growth, which raises the term premium. Watch the 30-year, not the equity bounce, for where this resolves.

3. Earnings Spotlight

Microsoft and Meta reported on the same night and told opposite stories. That contrast is the clearest evidence yet for how this market now prices AI.

4. Pre-Market Movers: Upside

5. Pre-Market Movers: Downside

6. Sector Rotation Radar

[ANALYSIS] The rotation has reached its cleanest form. The market now sorts AI names into two bins on a single question: does the spending show a return yet? Microsoft, yes, bought. Meta, not yet, sold. This is more discriminating than the sector-level rotation of two weeks ago and more precise than the compute-versus-infrastructure split of yesterday. It is company-by-company adjudication of the same arms race. The overlay that complicates everything is the 30-year yield: even proven AI winners carry long-duration cash flows that a 5.23% long bond discounts more harshly. The bull case needs both Azure-style proof and a stable bond market, and it only has the first.

7. Today’s Session Playbook

1. Trade the rebound, watch the 30-year. Microsoft’s Azure beat is a real catalyst and the bounce is justified, but the 30-year at 5.23% is the warning underneath it. If yields keep climbing, they will cap the rally regardless of earnings.
2. Top story: the invoice, settled. Microsoft +8% for proven monetization, Meta −8.5% for a 91% cash-flow collapse. Four weeks of this thesis resolved in one night, and the rule is now explicit: show the return, or get sold.
3. S&P 500 levels: Wednesday closed 7,316.15, breaking the 7,400 shelf decisively. Futures point to an open back near 7,360. Resistance is 7,400, then 7,428; support is 7,300, then 7,250. The Nasdaq is already in a technical correction, more than 10% off its high, which frames the bounce as a rebound within a downtrend until proven otherwise.
4. Event risk: the bond market and Apple/Amazon. A further leg up in the 30-year would undo the equity bounce on its own. Tonight, Amazon’s AWS growth either confirms or breaks the Microsoft-led relief, and Apple’s defensive profile gets its test. Two more binary catalysts before the week is done.
5. Under-the-radar: the stagflation signal. GDP at 1.5% with core PCE at 3.4% is the quietest and most important datapoint of the day. It puts the Fed in a vise and the long bond in revolt. Every equity move today is happening on top of a macro backdrop that just got measurably worse.

Four weeks ago this brief proposed a single organizing idea: the market had stopped paying for AI spending and started paying for contracted certainty, for a visible return on the capital being deployed. Every week refined it. Alphabet fell on record cloud growth because its capex went to $205B on faith. SK Hynix grew profit 557% and fell because the beat was not enough. The thesis kept working, but it was always a hypothesis about how the market would treat the biggest test of all, the hyperscalers reporting their own capex against their own returns. Wednesday night was that test, and it produced the cleanest result imaginable.

Microsoft and Meta reported within an hour of each other, in the same arms race, spending on the same AI future. Microsoft grew Azure 43%, held its capex steady, showed the monetization, and rose 8%. Meta grew revenue 28% but let free cash flow fall 91% to $784M against $31B of quarterly capex, could not show the return, and fell 8.5%. There is no cleaner controlled experiment in markets: same night, same sector, same thesis, opposite outcomes, and the only variable that differed was whether the spending had a visible payoff. The invoice thesis is no longer a hypothesis. It is the operating rule of this tape, and the market applied it with precision.

But the settlement of one question exposed a larger one, and intellectual honesty requires elevating it above the earnings story. While the market adjudicated AI winners and losers, the bond market delivered a verdict on everything at once. Warsh held rates and then refused, flatly, to signal any tolerance for above-target inflation, and the 30-year yield leapt to 5.23%, its highest since 2007. This morning’s data explained why the bond market is afraid: GDP slowed to 1.5% while core PCE held at 3.4%. That is the stagflationary box, weak growth and sticky prices, and it is the one configuration in which both stocks and bonds can be wrong at the same time. A 5.23% long bond raises the discount rate on every equity, including the Microsofts that just proved their AI economics work.

So the edge going into tonight and next week is a layered one. On the earnings axis, the rule is settled: watch AWS at Amazon for whether Microsoft’s cloud proof generalizes, and treat any AI capex without a matching return as a sell, because the market now does. But the axis that matters more is the macro one. The equity rebound is real and Microsoft earned it, yet it is happening in defiance of a bond market that is repricing long-run inflation and growth in real time. The tell for the second half is not which hyperscaler wins the AI race. It is whether the 30-year yield keeps climbing. If it does, it will eventually cap even the proven winners, because in a stagflationary regime the discount rate does what no earnings beat can undo. Trade the rebound with one eye on the long bond, and size for the possibility that this week’s real story was never the AI invoice at all, but the return of the bond vigilante.