The AI Capex Litmus Test: Microsoft and Amazon Pass, Meta and Apple Don't
A divided Fed, a $450 billion single-day rally, and a new standard for AI spending
The new capex litmus test
The week of July 28-31 delivered the clearest verdict yet on the AI capex cycle — not through a single earnings beat, but through a sharp and largely instantaneous bifurcation of the market’s judgment. Microsoft added roughly $450 billion in market capitalization in a single session after reporting Azure growth of 43% and a commercial remaining performance obligation (RPO) backlog of $678 billion[1]. Amazon followed by becoming the first company to surpass $200 billion in quarterly revenue ($200.61 billion), with AWS growing 37% — its fastest pace since 2021[1].
The contrast was stark. Meta’s earnings report included a 14% miss on EPS and free cash flow falling to $784 million, even as capex was raised to $130-$145 billion[1]. Alphabet raised capex to $195-$205 billion without a clear demand narrative[1]. Both were punished. Apple posted record revenue of $109.42 billion but fell on a Services miss ($30.74 billion vs. $31.22 billion expected) and weakness in Greater China ($18.8 billion vs. $19.6 billion expected)[1].
The market’s message, distilled: spend all you want on AI infrastructure — but show us the backlog, the revenue, and the customer demand first. The bar has moved from narrative to proof.
Friday’s close and the weekly scoreboard
| Index / ETF | Friday Close | Friday Change | Weekly Move |
|---|---|---|---|
| S&P 500 (SPY) | $747.03 | +0.72% | +1.03% |
| Nasdaq 100 (QQQ) | $687.99 | +0.65% | +1.77% |
| Dow Jones (DIA) | $524.32 | +0.54% | +0.53% |
| Russell 2000 (IWM) | $291.20 | -0.48% | -0.57% |
| Energy (XLE) | $59.55 | +1.00% | +2.04% |
| Tech (XLK) | $175.35 | -0.22% | ~+5% (Thu rally) |
| Gold (GLD) | $371.54 | -1.49% | — |
| 20+ Y Treasuries (TLT) | $82.25 | -0.66% | — |
Quotes as of 16:00 ET, July 31, 2026[2]. Weekly figures compiled from market recaps[3].
The shape of the week matters more than the headline number. Wednesday’s FOMC day was the low close for all four major indices. From that Wednesday trough, the S&P 500 added roughly 2.37% in two sessions, powered by Microsoft’s record-setting rally on Thursday and Amazon’s earnings beat on Friday[3]. The recovery was narrow: the Russell 2000 finished the week down even as large-cap tech surged, and the equal-weight S&P 500 had been hitting all-time highs on the Monday-Tuesday rotation into healthcare and staples[1].
The megacap scorecard
The five largest tech firms all reported within a four-day window, and the dispersion was extreme:
| Stock | Fri Close | Fri Move | Key Earnings Signal |
|---|---|---|---|
| MSFT | $464.72 | +3.02% | Azure +43%, RPO $678B, Copilot 30M seats |
| AMZN | $271.58 | +15.32% | AWS +37%, first $200B quarterly revenue |
| GOOGL | $356.13 | +6.73% | Recovering from prior week’s -15% |
| NVDA | $200.75 | +2.93% | Chip demand intact |
| META | $556.71 | +3.28% | EPS miss 14%, FCF $784M, capex raised |
| AAPL | $308.91 | -7.35% | Record revenue but Services & China miss |
Prices as of 16:00 ET close[4]. Post-market prints drifted slightly lower across the board[4].
A divided Fed holds the line
The FOMC voted 9-3 on July 29 to hold the federal funds rate in a 3.5%-3.75% range — the fifth consecutive meeting with no change[5]. Three regional presidents — Beth Hammack (Cleveland), Neel Kashkari (Minneapolis), and Lorie Logan (Dallas) — dissented in favor of tighter policy, marking the widest split in years[5].
The macro backdrop explains the tension. CPI inflation stands at 3.46% year-over-year, above the Fed’s 2% target for more than five years[6]. The 10-year Treasury yield sits at 4.68%, up 24 basis points month-over-month[6]. Yet unemployment is low at 4.2%, real GDP is growing at 2.1% YoY, and high-yield credit spreads remain tight at 2.84%[6]. The consumer, however, is unhappy: the University of Michigan sentiment index registered 49.5, down 18.45% year-over-year[6].
Fed-funds futures now price a 64% probability of a rate hike in September[3] — not a cut. Chairman Kevin Warsh, in his second FOMC meeting, signaled less forward guidance, leaving the market to read the data directly[5].
The closest historical macro analogs are mid-2006 and October 2007[6] — late-cycle periods when the Fed was holding or tightening, inflation was sticky above target, and the yield curve was positive but flattening. In 2006 the economy kept expanding; by late 2007 it did not. The difference between those two outcomes is the question hanging over everything right now.
The semiconductor signal
While U.S. megacaps sorted themselves into winners and losers, South Korea’s KOSPI recorded its largest single-day gain ever — 17.91% — with SK Hynix up 30% and Samsung up 27%[1]. The catalyst was Amazon’s explicit confidence in AI semiconductor demand through 2027, coupled with a 199.5% surge in June exports to $44.8 billion[1].
This matters beyond Korea. If hyperscaler demand for AI compute is visible enough to trigger the KOSPI’s biggest day in history, the semiconductor supply chain is receiving a demand signal that extends well beyond any single quarterly print. It is the same signal embedded in Microsoft’s $678 billion backlog and Amazon’s capacity-constrained AWS growth — just observed from a different vantage point in the supply chain.
What to watch next
The week of August 3-7 carries the earnings baton forward against a more macro-heavy backdrop[7]:
- Palantir (Monday) — the first read on AI software monetization outside the hyperscalers
- AMD (Tuesday) — a test of whether the GPU and accelerator demand that lifted Nvidia extends to the second source
- SanDisk (Wednesday) — storage demand tied to AI infrastructure build-out
- SpaceX (debut earnings) — a first public look at the space-economy narrative’s financials
- July NFP jobs report (Friday) — the single most important data point for a Fed that just split 9-3 and faces a 64% market-implied probability of a September hike
The base-rate question for the coming week: does the proof-of-demand bifurcation from megacap earnings extend to the next tier of AI-exposed names, or was it a one-off driven by the size and visibility of Microsoft and Amazon? If Palantir and AMD can point to backlog and validated demand, the framework holds. If they cannot, the market’s new capex test will look less like a durable standard and more like a week-specific verdict on two companies.
The larger question — whether the economy is in a mid-2006 expansion-with-sticky-inflation or a late-2007 pre-recession — will not be answered by earnings alone. Friday’s jobs report, against a 49.5 consumer sentiment reading and a divided FOMC, will carry more weight than usual.
FN2 Research provides market commentary and education, not personalized investment advice. Index and ETF prices cited from end-of-day data as of July 31, 2026.