Published: July 28, 2026 | Reading Time: ~12 minutes | Channel: Business
Let me give you the number that matters most this week: $725 billion. That's what Microsoft, Meta, Amazon, and Alphabet are on track to spend on AI infrastructure in 2026 — a 77% jump from the $410 billion they burned through last year.¹
And last week, Alphabet became the canary in that very expensive coal mine.
Google's parent company delivered what should have been a victory lap. Q2 revenue hit $119.8 billion, up 25% year over year. Google Cloud exploded 82% higher to $24.8 billion. Search and advertising grew 17%. By any normal standard, these are the kind of numbers that send champagne corks through boardroom ceilings.²
Instead, Alphabet shares plunged more than 7% — their worst single-day drubbing in over a year. The reason? CFO Anat Ashkenazi raised the company's 2026 capital expenditure guidance to $195–$205 billion and revealed that quarterly free cash flow had turned negative for the first time since the company's 2004 IPO.³
The market's message was brutal and unambiguous: we don't care how fast your cloud business is growing. We want to know when the spending stops — and when the profits start.
This week, three more tech titans step into the same buzzsaw. Microsoft and Meta report Wednesday. Apple and Amazon follow Thursday. By Friday afternoon, we'll know whether Alphabet was an isolated warning shot or the opening salvo of a full-scale investor revolt against AI spending.
Let's get the details right, because the story is more nuanced than the headline.
Alphabet didn't just beat estimates — it demolished them. Total revenue of $119.8 billion cleared consensus by nearly $3 billion. Google Cloud's 82% surge to $24.8 billion wasn't just a beat; it was an acceleration from what was already breakneck growth. Operating income reached $34.7 billion.⁴
And yet the stock cratered. Why?
Because on the same earnings call, Alphabet dropped two bombs:
Bomb #1: Full-year 2026 capex guidance raised to $195–$205 billion, up from the prior $180–$190 billion range. That's a $15 billion increase at the midpoint — and management explicitly said spending would scale further in 2027.
Bomb #2: Quarterly free cash flow landed at negative $5.9 billion. For a company that generated $69 billion in free cash flow just last year, the reversal was staggering.
Jason Lemire, chief investment officer at Bold Wealth Partners, captured the sentiment shift perfectly: "People are really focused on capex, obsessed with it. It used to be the more the better, but now it is the less the better. We're seeing capital raises, negative cash flows, rising debt. All that adds risk to the picture."⁵
Translation: the "spend whatever it takes" era is over.
Here's what each of the Big Four hyperscalers is committing to AI infrastructure this year:
| Company | 2026 Capex (Est.) | 2025 Capex | YoY Increase | Reports |
|---|---|---|---|---|
| Amazon | ~$200B | Not disclosed separately | — | Thurs, July 30 |
| Microsoft | ~$190B | ~$118B | +61% | Wed, July 29 |
| Alphabet | $195–$205B | ~$105B | ~+90% | Already reported |
| Meta | $125–$145B | $60–$65B | ~+115% | Wed, July 29 |
| TOTAL | ~$725B | ~$410B | +77% | — |
And Goldman Sachs now projects cumulative hyperscaler capex of $5.3 trillion from 2025 through 2030 — a figure that was $4.5 trillion before Q1 earnings season began.⁶
Let that sink in. In the span of one quarter, the projected 6-year AI infrastructure bill grew by $800 billion. That's roughly the GDP of Switzerland. Added in 90 days. On top of numbers that were already eye-watering.
If you want to know how deep this goes, look at what happened in Seoul on Tuesday morning.
South Korea's Kospi index tumbled more than 10% as memory chipmakers SK Hynix and Samsung Electronics cratered 14% and 13% respectively.⁷ The trigger? Reports that Nvidia is exploring a $250 billion funding backstop for OpenAI — a move that would further intertwine the two companies and intensify worries about "circular financing" in the AI ecosystem.
Here's the uncomfortable question nobody on CNBC wants to ask out loud: what happens if Nvidia is effectively lending money to its own customers so those customers can keep buying Nvidia chips?
The Philadelphia Semiconductor Index (SOX) is already down 17% in July, on pace for its worst month since June 2022. After soaring 101% through the first half of the year, the 30-member chip index has recorded 17 daily moves of 5% or more in 2026 — matching the most since 2008. By comparison, the S&P 500 has had exactly zero.⁸
Bold Wealth's Lemire again: "There is going to be an AI winter at some point. When you look at how exceptional margins are — especially in memory — well, it is impossible to maintain those over a long timeframe. At some point, we will see margin compression and valuation compression, and that will have a huge impact on the market."⁹
Here's where I part ways with the panic.
Yes, $725 billion is an incomprehensible number. Yes, Alphabet's negative free cash flow is genuinely concerning. Yes, the semiconductor selloff is real and may have further to run.
But the narrative that "Big Tech is burning cash with nothing to show for it" is lazy and wrong. Here's what the market is missing:
1. Google Cloud grew 82%. That's not a rounding error. That's a business line that barely existed as a meaningful entity five years ago and is now generating revenue at a $100 billion annual run rate. AI infrastructure spending is what made that possible.
**2. Microsoft's Copilot enterprise seats exceeded 20 million in fiscal Q3 — with seat additions up 250% year over year.**¹⁰ That's real revenue from real customers choosing to pay real money for AI tools.
3. Microsoft's MAI models are already slashing costs. When Microsoft deployed its own MAI-Image-2.5-Pro model in Bing Image Creator, GPU costs dropped 84% compared to OpenAI's GPT-Image-2. MAI-Voice-2-Flash achieved 89% cost reductions in Dynamics 365.¹¹ The capex isn't just spending — it's building infrastructure that eventually replaces expensive third-party licensing.
4. Meta's AI advertising tools are driving measurable results. Ad impressions grew 19% and average ad prices rose 12% in Q1 — both metrics that directly flow to the bottom line.¹²
The problem isn't that AI spending produces no returns. The problem is that the market wants every dollar of capex to produce two dollars of revenue by next Tuesday.
Here's the most fascinating subplot of the week, and the one nobody's talking about enough.
Apple — which has spent almost nothing on AI infrastructure compared to its peers — is the best-performing Magnificent Seven stock of 2026. It's up 23% year to date and 15% in July alone. Microsoft is down 21%. Meta is down 9.8%. The Mag Seven index as a whole is down 3.7%.¹³
Apple's strategy? Partner with model developers instead of building everything in-house. Let Google, OpenAI, and others spend a trillion dollars on GPUs and data centers. Then license their models, integrate them into Apple Intelligence, and collect a 30% cut on the App Store revenue they generate.
Tim Cook called the memory pricing situation a "hundred-year flood" — something he'd never seen in 40 years in the industry.¹⁴ And he's right. The same HBM (high-bandwidth memory) supply crunch that's forcing Microsoft and Meta to pay premium prices for AI accelerators is hitting Apple from the other direction. The 12GB DRAM package that cost Apple $39 per iPhone 17 Pro is projected to cost $145 for the iPhone 18 Pro — a 272% increase.¹⁵
Apple already responded by raising Mac and iPad prices up to $300 on June 25. iPhone 18 pricing will come this fall, and analyst estimates for the increase range from $50 (JPMorgan) to $270 (TechInsights' full margin-preservation scenario).¹⁶
But here's the key difference: Apple's memory cost problem is a supply-chain issue, not a failed-investment issue. The company didn't spend $200 billion building data centers that might or might not pay off. It's dealing with input cost inflation — a problem every manufacturer faces. The market is rewarding that capital discipline.
This earnings week isn't just a spectacle — it's a stress test that will create genuine opportunity for disciplined investors. Here are five concrete plays to consider:
1. Watch Azure growth, not Microsoft EPS. Microsoft reports Wednesday after close. The number that matters is Azure's year-over-year constant-currency growth rate. Management guided 39–40%. The "sell-off floor" is approximately 36%. If Azure prints above 40%, buy the dip on any capex-related selloff. If it prints below 36%, the growth story is genuinely cracking.¹⁷
2. Meta's capex guidance is the real earnings number. Revenue will be strong — consensus is around $60.3 billion. But if Meta raises its 2026 capex range above the current $125–$145 billion, the stock will get punished regardless of ad revenue growth. If they hold the line or signal a spending peak, the stock could rip.¹⁸
3. Apple is the safest harbor — but at a price. Apple reports Thursday, and this is Tim Cook's final earnings call as CEO before John Ternus takes over September 1.¹⁹ Hardware gross margin — currently 38.7% — is the tell. If memory costs are compressing margins faster than Services (which hit $26.3 billion in Q2) can offset them, even Apple isn't safe.
4. The semiconductor trade needs patience. The SOX is down 17% in July. Nvidia is exploring a $250 billion OpenAI backstop. SK Hynix reports tonight. The memory trade isn't dead — global AI demand is real — but the easy money has been made. Wait for the panic to exhaust itself before adding. The SOX has 17 moves of 5%+ this year; there will be better entry points.
5. The Fed meeting adds another variable. The FOMC begins its two-day meeting today. Traders are split between hold and hike — one of the most uncertain policy decisions in years. If the Fed hikes, the capex narrative gets even uglier as higher rates increase the cost of funding all that infrastructure. If they hold, the pressure eases slightly. Consumer confidence data at 10 AM ET today will provide the last signal before Wednesday's decision.²⁰
1. The Circular Financing Trap. If Nvidia is bankrolling OpenAI so OpenAI can buy Nvidia chips, we're not witnessing a technological revolution — we're witnessing a balance-sheet ponzi scheme. When one company's revenue depends on lending money to customers who use that money to pay the same company, you have a circular dependency that unravels violently if any link breaks. The $250 billion backstop figure is not confirmed, but the fact that it's even being discussed should terrify you.
2. The Memory Cartel Problem. Three companies — Samsung, SK Hynix, and Micron — control essentially all high-bandwidth memory production. They're rationally prioritizing HBM (3–5x revenue per wafer) over consumer DRAM. But that rational decision is creating a supply crisis for every device maker on Earth. If Apple has to raise iPhone prices by $270 to preserve margins, demand destruction becomes a real risk. The memory manufacturers are printing money today at the expense of their broader customer ecosystem.
3. The Microsoft-OpenAI Divorce Isn't Priced In. In April, Microsoft and OpenAI formally ended their exclusive partnership. At Build 2026, Microsoft unveiled seven in-house MAI models built without any distillation from OpenAI.²¹ The MAI cost savings are impressive — but they also mean Microsoft is now competing with the company it invested $13 billion in. If OpenAI's next model underwhelms while Microsoft's in-house models accelerate, Microsoft's OpenAI stake becomes an increasingly awkward asset.
4. The 2027 Capex Cliff. Alphabet already signaled that capex will scale further in 2027. Morgan Stanley projects Microsoft's 2027 capex could hit $262 billion.²² At some point — and we may already be there — the market stops treating these investments as "growth capex" and starts treating them as "maintenance capex" that merely preserves competitive position. That re-rating alone could compress valuations by 30–40%.
The AI spending panic is real, justified in parts, and absolutely creating opportunity for anyone who can separate signal from noise. Alphabet's 82% cloud growth proves the spending IS producing returns. Microsoft's 250% Copilot seat growth proves enterprises ARE buying. But the market has shifted from "tell me how much you're spending" to "show me the receipt" — and that's a healthy correction, not a crisis.
The single most important number this week is Microsoft's Azure growth rate on Wednesday. If it holds above 40%, the AI trade is alive and the selloff is a buying opportunity. If it cracks below 36%, batten the hatches.
Either way: don't panic. Do the math. And remember that Apple — the company that spent the least — is winning the most.
Yahoo Finance / Brian Sozzi — Meta, Microsoft, Amazon, and Alphabet to spend $725B combined in 2026 on capex, up 77% from $410B. https://finance.yahoo.com/sectors/technology/article/meta-microsoft-amazon-and-alphabet-are-about-to-spend-a-shocking-amount-of-money-to-dominate-the-ai-era-115359575.html
Fortune / Bloomberg — Jeran Wittenstein, Ryan Vlastelica — Alphabet Q2 revenue $119.8B (+25%), Google Cloud +82% to $24.8B, Mag Seven index down 3.7% YTD. https://fortune.com/2026/07/26/big-tech-earnings-meta-microsoft-apple-amazon-market-revolt-ai-spending/
TOPONE Markets — Alphabet capex raised to $195–$205B, FCF turned negative -$5.9B, stock down 7.38%. https://www.top1markets.com/news/alphabet-stock-drop-q2-2026-earnings-capex-ai-spend
TOPONE Markets — Alphabet Q2 operating income $34.7B, Search & Ad revenue $63.3B (+17%), YouTube $11.1B (+13%). https://www.top1markets.com/news/alphabet-stock-drop-q2-2026-earnings-capex-ai-spend
Fortune / Bloomberg — Jason Lemire (Bold Wealth Partners) on capex obsession and investor sentiment shift. https://fortune.com/2026/07/26/big-tech-earnings-meta-microsoft-apple-amazon-market-revolt-ai-spending/
Yahoo Finance / Brian Sozzi — Goldman Sachs $5.3T cumulative hyperscaler capex estimate 2025–2030, up from $4.5T. https://finance.yahoo.com/sectors/technology/article/meta-microsoft-amazon-and-alphabet-are-about-to-spend-a-shocking-amount-of-money-to-dominate-the-ai-era-115359575.html
Yahoo Finance Live / Grace O'Donnell — Kospi -10%, SK Hynix -14%, Samsung -13%, Nasdaq futures down nearly 1%. https://finance.yahoo.com/markets/live/stock-market-today-tuesday-july-28-dow-sp-500-nasdaq-082832371.html
Fortune / Bloomberg — SOX down 17% in July, 17 daily moves of 5%+ matching most since 2008. https://fortune.com/2026/07/26/big-tech-earnings-meta-microsoft-apple-amazon-market-revolt-ai-spending/
Fortune / Bloomberg — Lemire on AI winter risk, margin compression, valuation compression. https://fortune.com/2026/07/26/big-tech-earnings-meta-microsoft-apple-amazon-market-revolt-ai-spending/
TechTimes / Daniel Butler — Microsoft Copilot enterprise seats 20M+, seat additions +250% YoY, commercial backlog doubled. https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
TechTimes / Daniel Butler — MAI-Image-2.5-Pro 84% GPU cost reduction vs OpenAI, MAI-Voice-2-Flash 89% reduction. https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
TechTimes / Daniel Butler — Meta Q1 ad impressions +19%, average price +12%, revenue $56.3B (+33%). https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
Fortune / Bloomberg — Apple +23% YTD, +15% July; Microsoft -21% YTD; Meta -9.8% YTD; Mag Seven -3.7% YTD. https://fortune.com/2026/07/26/big-tech-earnings-meta-microsoft-apple-amazon-market-revolt-ai-spending/
TechTimes / Daniel Butler — Tim Cook "hundred-year flood" quote on memory pricing. https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
TechTimes / Daniel Butler — iPhone DRAM cost $39 → $145 (272% increase), NAND 256GB $13 → $51. https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
TechTimes / Daniel Butler — Mac/iPad price hikes up to $300 on June 25, iPhone estimates $50 (JPMorgan) to $270 (TechInsights). https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
TechTimes / Daniel Butler — Azure growth guidance 39-40%, informal sell-off floor ~36%, fiscal Q4 EPS $4.22-4.24. https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
TechTimes / Daniel Butler — Meta Q2 consensus $60.3B revenue, $7.19 EPS, $125-145B FY capex range. https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
TechTimes / Daniel Butler — Tim Cook's final earnings call, John Ternus CEO from September 1. https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
Yahoo Finance Live / Grace O'Donnell — Fed meeting starts today, consumer confidence at 10 AM ET. https://finance.yahoo.com/markets/live/stock-market-today-tuesday-july-28-dow-sp-500-nasdaq-082832371.html
TechTimes / Daniel Butler — Microsoft-OpenAI partnership ended April 2026, 7 MAI models at Build 2026, no OpenAI distillation. https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
TechTimes / Daniel Butler — Morgan Stanley projects Microsoft 2027 capex at $262B if trajectory holds. https://www.techtimes.com/articles/321695/20260727/big-tech-earnings-week-one-memory-crisis-four-different-balance-sheets.htm
All claims verified against Gold-tier (Bloomberg/Fortune, Yahoo Finance) and Silver-tier (TechTimes, TOPONE Markets) sources. Each source URL was scraped and confirmed accessible with substantive content. Last verified: July 28, 2026.
The market doesn't pay you for agreeing with it. It pays you for being right before it realizes it was wrong. 🎯
