It was one of the most data-heavy days in recent market history. Google, Amazon, Meta, and Microsoft — companies that together represent just under 20% of the total market cap of the S&P 500 — all reported earnings on the same day. Throw in a Federal Reserve interest rate decision and a Jerome Powell speech, and you had a full-on financial market spectacle. The Fed held rates at 3.5–3.75%, meeting expectations but disappointing anyone hoping for a cut. But the real story was what happened after 4 p.m. Eastern, when the big tech earnings quad-kill landed all at once.
What Happened With Big Tech Earnings Today?
The central question running through all four earnings reports was the same: can these companies turn massive AI capital expenditure into durable, growing revenue before depreciation catches up? Every hyperscaler is spending at least $100 billion. Amazon has guided toward $200 billion in capex for 2026 — the first time any company has ever posted a number that large. The market wanted proof that the spending wasn't just a bet, but a business.
00:45
The earnings schedule breakdown: Fed decision, Powell speech, and four mega-cap reports all in one day
Watch at 00:45 →
Going into the reports, analyst notes from firms like Semi-Analysis were already flagging optimism. The expectation was that hyperscaler capex guidance would be revised upward, cloud revenue was accelerating, and companies were seeing positive ROI on cloud investments. That backdrop set the stage for what became a very closely watched set of results.
Financial performance in legacy areas has actually been strong across the board. Google Search is growing. Amazon's core e-commerce business is chugging along. Microsoft 365 seat counts are expanding. But the big question — the one Wall Street was really asking — is whether AI infrastructure revenue is durable, and what the margins look like in what increasingly resembles a railroad or oil business rather than a high-margin software subscription play.
How Much Is Amazon Spending on AI Capex?
Amazon is the capex headline of the moment. The company guided toward $200 billion in capital expenditure for 2026 — a number so large it's the first time any company has ever printed it. For context, Amazon's operating income last quarter was $25 billion and free cash flow came in at $11.2 billion. That free cash flow number was already down year-over-year due to increased AI spending.
The logic behind the spending is straightforward: if AWS accelerates, all that capex looks like smart, ahead-of-the-curve capacity buying. Being GPU-rich and compute-rich at exactly the moment when demand is surging is the dream scenario. The bear case is that you're drawing down cash at an accelerating rate with depreciation timelines that don't match your revenue ramp. That's the knife-edge Amazon is walking.
08:20
Amazon's $200B capex guidance and what it signals about GPU demand
Watch at 08:20 →
Is AWS Revenue Growth Accelerating in 2025?
AWS grew at 24% in Q4, with net sales for Amazon overall up 14%. The sneakily massive Amazon Ads business grew 23% to $21.3 billion — the company is approaching $100 billion a year in ad revenue alone, which is an enormous cash engine to fund all that AI capex.
But the market wants more from AWS specifically. Consensus heading into the report had AWS revenue around $36.7 billion with mid-20% growth. The real excitement would be a reacceleration into the 30% range — anything starting with a three. That would signal that enterprise cloud demand is genuinely inflecting upward, not just holding steady. Strong AWS acceleration is the single number that justifies the $200 billion capex bet most cleanly.
How Is Meta Actually Making Money From AI?
Meta is arguably the most interesting AI story in this group because it doesn't have a diffusion problem. Every other company has to figure out how to get businesses or consumers to actually adopt and pay for AI features. Meta doesn't. When they improve a model that places ads more effectively, they A/B test it, roll it out across Instagram, Facebook, and Threads, and users never even notice — they just see slightly better ads.
14:35
Meta's ad flywheel: how AI model improvements translate directly to ad revenue without a diffusion lag
Watch at 14:35 →
The numbers back this up. Meta reported 3.58 billion daily active users, revenue growth of 24% to nearly $60 billion last quarter, ad impressions up 18%, and average price per ad up 6%. Family of apps operating income hit $30.8 billion. Reality Labs lost $6 billion, which at this scale feels almost quaint — a side bet on the future of devices that the ad machine easily absorbs.
Capex was $72.2 billion last year. The guide for this year is $115–135 billion, nearly a doubling. The key question going into the report: does the AI improvement in ad placement continue to compound? Because at Meta, a breakthrough in reasoning or model quality translates to better ad performance almost immediately — no enterprise sales cycle, no change management, no security review. Just better ads, better margins, faster.
Expectations implied revenue growth of around 31%. The more AI improves Meta's ad targeting, the more that number could surprise to the upside.
Will AI Overviews Kill Google's Search Ad Revenue?
Google has arguably the most fully integrated AI stack of any company on earth. Consumer distribution, model training through DeepMind, custom chips via TPU, and deployment surfaces across Search, YouTube, Google Workspace, Android, and Cloud. The flywheel theoretically spins faster than anyone else's.
But the core investor anxiety around Google is unit economics: do AI Overviews and Gemini expand or compress the search ad revenue model? If users get direct answers from an AI summary instead of clicking through to ads, does that reduce ad inventory and revenue? Or does better search quality increase engagement and ultimately ad ROI?
Google Search has continued growing, which is a positive signal. And even if core search growth moderates, Google Cloud is an accelerating business with significant upside. The question is whether AI features are expanding usage and ad performance or pulling the rug out from under the business model that funds everything else. Earnings gave the market a fresh data point on exactly that tension.
What Is Microsoft's $625B RPO and Why Does It Matter?
Microsoft came into this earnings cycle off a very strong quarter. Revenue was up 17%, Microsoft Cloud grew at 26%, and Azure specifically grew at 39%. But the single most eyebrow-raising number from last quarter was the Remaining Performance Obligations (RPO) figure of $625 billion — up 110% year-over-year.
RPO represents signed contracts for future compute and cloud services. About 45% of that number comes from OpenAI, but there are many other long-term compute contracts in there. What it signals is that the capex Microsoft is spending isn't speculative — companies have already signed contracts committing to pay for that capacity. The revenue is, in a meaningful sense, pre-sold.
Microsoft also has the cleanest read on enterprise AI monetization of any company in the world. Every major American business runs on Microsoft infrastructure. So the data points that matter most out of their earnings — Azure growth, cloud gross margins, Copilot adoption rates, M365 seat growth, GitHub Copilot momentum — paint a real-time picture of how AI is actually diffusing through the global economy, not just the hype cycle.
The M365 seat growth number is particularly interesting as a bellwether. Does an AI agent need its own seat? Does one seat now serve twenty agents? These longer-term structural questions are starting to show early signals in the data.
The Red vs Blue Button Thought Experiment Explained
In between earnings coverage, the show dug into a viral thought experiment that originally went big through Tim Urban's Wait But Why blog (24 million views) and then got amplified by Mr. Beast (10 million views on a similar prompt). Here's the scenario:
- Everyone in the world votes red or blue.
- If more than 50% press blue, everyone survives.
- If fewer than 50% press blue, only the red button pressers survive.
The hyper-rational answer is red — no matter what anyone else does, pressing red guarantees your survival. But if everyone reasons that way, you hit under 50% blue, and everyone who pressed blue dies. The cooperative answer is blue — you only need 51% of people to coordinate, and if it works, literally no one dies.
In Tim Urban's poll, 58% went blue. In Mr. Beast's version, 55.7% went blue. In both cases, humanity survived the hypothetical. A follow-up survey of 14,000 people found that blue button selection correlated strongly with self-reported truth-telling. The chat during the show leaned red — which, the hosts noted, says something about the audience's risk calculus or distrust of coordination. Either way, it's a surprisingly clean frame for thinking about collective action problems, trust, and rational self-interest — themes that aren't entirely unrelated to how AI infrastructure investment actually works at a macro level.
