AI Capex Scorecard: What Earnings Week Actually Showed

The AI capex scorecard for July 2026's earnings week is a study in opposite verdicts. Five Big Tech companies reported in nine days — Alphabet on July 22, Microsoft and Meta on July 29, Amazon and Apple on July 30 — and every one of them posted double-digit year-over-year revenue growth. The market rewarded two, punished three, and the dividing line was not the size of the AI spend.

What’s at stake is the single most-watched question in equities right now: is the hundreds of billions of dollars Big Tech is pouring into AI infrastructure a durable investment or an escalating liability? This week produced the clearest answer yet — and it isn’t the lazy “market punishes AI overspending” story that framed some of the coverage. The four infrastructure owners on this scorecard all raised or sustained elevated capital spending. The reactions still split.

This post runs all five reports through one lens: what each company spent or guided, what AI revenue it could point to, how the stock reacted, and what actually drove the verdict. Every earnings outlet covered its own company; nobody built the side-by-side. That table is below, along with what the pattern means for anyone buying ads, running cloud workloads, or planning an AI budget of their own.

Here is the week in a single view — capex position, AI-revenue proof point, market reaction, and the causal driver behind each verdict. Capex figures are labeled by basis (guided vs. actual vs. trailing-twelve-month) because mixing them is exactly how bad takes get written. Market reactions are quoted as ranges: reported percentages vary by outlet and by measurement window for all five companies, so no single precise number is definitive.

* Share reactions are ranges across outlets and measurement windows (immediate after-hours vs. next-day trade), compiled from company IR releases and coverage by CNBC, Fortune, TradingKey, SiliconANGLE, and 24/7 Wall St, July 29–30, 2026. The five-company side-by-side and the “what drove the verdict” column are our analysis; each underlying figure is sourced to the company’s own release or call, with press corroboration for the market-reaction cells.

Microsoft’s fiscal Q4 (quarter ended June 30, reported July 29) was the week’s cleanest bull case. Per the company’s investor-relations release, total revenue hit $90.0B, up 18% YoY, with operating income of $40.6B (+18% YoY) and GAAP net income of $35.8B (+31% YoY). Microsoft Cloud reached $59.3B, up 27% YoY. The headline inside the headline: Azure revenue growth accelerated to 43% YoY, and Azure’s trailing-twelve-month revenue crossed $100B for the first time in company history.

The written release confirmed the AI proof point that mattered most: Microsoft 365 Copilot paid seats reached over 30 million, up from just over 20 million in April 2026 — roughly 50% seat growth in a single quarter. Contracted backlog rose $51B in the quarter, from $627B to $678B. Worth an attribution note, because it circulated widely during earnings week: the AI annualized run rate above $37B, described as up 123% YoY, is Microsoft’s Q3 FY26 disclosure from April 29, 2026 (the quarter ended March 31), not something the company re-disclosed on this July 29 call. Anyone quoting it against these Q4 numbers is mixing quarters.

The market’s response — a rise reported anywhere from +3% in the immediate after-hours print to roughly +8–9% by the next trading day, depending on the outlet and window — was among the week’s strongest rewards. One post-earnings analysis noted that of Microsoft, Meta, and Alphabet, Microsoft was the only one whose capex number moved in the direction the market wanted. We’d sharpen that: the capex number moved up, like everyone else’s. What moved in the right direction was the revenue attached to it.

Amazon’s Q2 (reported July 30) delivered its first $200B quarter — net sales of $200.6B, up 20% YoY, per the company’s investor-relations release. The AI story lives in AWS: net sales of $42.2B, up 37% YoY — the segment’s fastest growth in 18 quarters, against Wall Street’s ~31% growth expectation. AWS operating income reached $16.6B, up from $10.2B a year earlier, at a 39.4% operating margin. Advertising services added $19.8B, up 26% YoY. Shares rose roughly 8–9% in after-hours and next-day trading. We break the full report down in our companion post on Amazon’s $200B quarter.

One more discipline point: Amazon issued no full-year 2026 capex guidance. The only primary-confirmed spending figure is a trailing-twelve-month actual of $173.0B in capital expenditures, net of finance-lease proceeds. Pre-earnings headlines floating a “$200B capex plan” were speculation, not a confirmed guide — and the scorecard above labels Amazon’s number as a TTM actual for exactly that reason.

Meta’s Q2 (reported July 29) is the report that breaks the naive “good numbers up, bad numbers down” model. Revenue hit $60.8B, up 28% YoY, per the company’s earnings release, with advertising revenue of $59.4B, up 27% YoY — ad impressions up 14% YoY and average price per ad up 12% YoY. The ad machine is not the problem. The stock still fell roughly 9–10% in after-hours trading, with the slide continuing into July 30.

Three numbers did the damage. Net income fell 14% YoY to $15.8B and diluted EPS of $6.18 (down 13% YoY) missed consensus of roughly $7.13–7.14, weighed down by elevated legal and severance charges — full-year total expense guidance rose to $165–169B, incorporating a $2.4B legal charge. Quarterly capex hit $31.1B, and the full-year guide was raised to $130–145B, with the low end moving up from the prior $125–145B range. And free cash flow collapsed to $784M — down roughly 91% YoY from $8.5B a year earlier, even as operating cash flow grew 24.6% YoY to $31.9B. The cash is going into the ground as fast as it comes in.

CEO Mark Zuckerberg’s framing in the release was confident: “AI is accelerating our core business today, powering our next generation of products, and opening the door to entirely new enterprise opportunities. The results are already showing, and I’m optimistic about the potential ahead.” The market’s issue is that “accelerating our core business” is an internal-efficiency claim — there is no metered, external AI revenue line to watch grow. Asked about next year, the CFO declined to commit:

That is a defensible operating posture and a hard sell to a market pricing AI spend against AI revenue. Note that Microsoft withheld an FY27 figure too — it named roughly $175B for calendar 2026 and left next fiscal year at “grows year over year” — and rallied anyway. So the asymmetry isn’t about candor or disclosure. It’s that Microsoft’s spending comes attached to a consumption revenue line growing 43% YoY, and Meta’s would come attached to a promise.

Apple’s fiscal Q3 (quarter ended late June, reported July 30) was by its own description the strongest June quarter in company history: revenue of $109.4B, up 16% YoY, and diluted EPS of $2.02, up 29% YoY — figures that include a $0.11 favorable per-share impact from tariff refunds, with gross margin of 50.1% carrying roughly two points of tariff-refund benefit, per the company’s Newsroom release. “Today, Apple is proud to report our strongest June quarter ever, with double-digit revenue growth across iPhone, Mac and Services,” said CEO Tim Cook — on what was also his final earnings call before John Ternus takes over as CEO on September 1, 2026. Per press aggregation of the release and call (9to5Mac), iPhone revenue of roughly $54.3B (+22% YoY) and Mac at roughly $10.4B both beat estimates, while iPad (~$6.2B) and Services (~$30.7B) fell short.

The shares still fell — roughly 4–8%, depending on the trading session measured. The trigger was guidance: revenue growth of 9–11% for the next quarter, below the ~12% analysts expected, with management explicitly citing memory-chip cost inflation and advanced chip-packaging capacity constraints tied to broader AI-driven chip demand. Cook also acknowledged, per CNBC’s reporting of the call, that it is still early for Apple in terms of what heavy AI usage will mean for costs. The Gemini-powered Siri rebuild remains on track for a fall 2026 launch — covered in depth here. Our full read of the quarter is in the companion Apple post.

Stack the disclosures and the scale becomes the story. Three of the five companies put a number on forward capex for 2026: Microsoft at roughly $175B for calendar 2026 (with FY2027 guided only to grow year over year), Meta at $130–145B for FY2026, and Alphabet at $195–205B for FY2026 — raised from $180–190B a week earlier, as we covered in our Alphabet Q2 breakdown. Those three 2026 guides sum to $500–525B. Amazon issued no forward guide, but its trailing-twelve-month actual of $173.0B puts the four-company disclosed AI-era infrastructure bill at roughly $675–700B on an annualized basis — with the honest caveat that this mixes forward guides with a trailing actual, and that none of the four has quantified a full-year figure beyond 2026.

Where does it go? Mostly into chips, datacenters, and power — the supply chain we’ve been tracking all year, from the compute-diversification deals sitting behind these capex numbers to who’s actually financing the buildout. Alphabet’s Q2 also carried the week’s starkest cash-flow signal: operating cash flow of $39.1B could not cover the spending, producing negative free cash flow of $5.9B — the company’s first negative-FCF quarter since its 2004 IPO. Cloud revenue up 82% YoY with a $514B backlog bought Alphabet a milder penalty (a reported ~5% decline) than Meta’s, but not a pass.

Line the five verdicts up and the popular narratives fall apart. “The market is punishing AI overspending” fails immediately: the two biggest spenders by forward guide — Microsoft and Alphabet — got the week’s biggest reward and one of its milder penalties, respectively, while Apple, which barely owns AI infrastructure at all, was punished. “Beat and rise” fails too: Meta and Apple both beat on revenue and both fell. Capex direction doesn’t predict the verdict either — the four infrastructure owners all raised or sustained elevated spending, and the reactions still split.

The variable that does sort all five outcomes is what we’d call spend-to-revenue legibility: can the company point its AI capex at a metered, external, growing revenue line? Microsoft can (Azure, +43% YoY, TTM past $100B). Amazon can (AWS, +37% YoY, at a 39.4% operating margin). Meta cannot yet — its AI spend routes into internal ad-system efficiency and still-nascent business-AI bets, so the market fell back on what it could see: an EPS miss and a 91% YoY free-cash-flow decline. Apple’s AI story is a fee paid to a partner plus cost exposure to everyone else’s buildout. Alphabet sits in between: Cloud growth of 82% YoY is legible, but it couldn’t outweigh the optics of the first negative-FCF quarter since 2004.

Projecting forward, this framework makes the next two quarters easy to read. Meta’s stock recovery likely depends less on ad growth than on producing a number the market can meter — business-AI revenue, AI-attributable pricing gains, anything external and recurring. Microsoft and Amazon have effectively been handed permission to keep spending as long as Azure and AWS growth holds; the risk flips only if consumption growth decelerates while the guides stay up. And if Alphabet’s free cash flow stays negative for a second quarter, expect the “legible but unaffordable” framing to harden. The market didn’t vote against AI capex this week — it voted against AI capex it can’t measure.

Earnings week wasn’t just an investor story. The same numbers set the operating environment for anyone spending on these platforms.

The measurement point deserves emphasis. The market just demonstrated, at half-a-trillion-dollar scale, that unmetered AI spend gets punished regardless of how good the underlying business is. The same discipline applies to a marketing budget: if you’re increasing spend on AI-assisted campaigns, the attribution has to be in place first — that’s the core of our analytics and measurement work, and it’s why our paid media engagements start from metered outcomes rather than platform narratives. For teams making the build-vs-buy call on AI systems of their own, our AI transformation practice runs exactly this spend-to-revenue mapping before a dollar is committed.