NVIDIA, Microsoft, Amazon and Alphabet are repricing the AI ceiling

Five Tech Giants Just Rewrote the Rules of Scale

Last week was one of those rare stretches where you sit back after Friday's close and think: the map just changed.

Five of the biggest companies on earth reported earnings within days of each other. Tesla (NASDAQ:TSLA), NVIDIA (NASDAQ:NVDA), Alphabet (NASDAQ:GOOG), Amazon (NASDAQ:AMZN), and Microsoft (NASDAQ:MSFT). Combined market cap somewhere north of $15 trillion. And if you read the transcripts rather than the headlines, one theme kept surfacing: these companies are not managing their current businesses. They are constructing the next ones.

Start with NVIDIA. Jensen Huang is approaching a $300 billion annualized revenue pace. Data center alone hit $62 billion in a single quarter, up 75% year-over-year. The thing most analysts will spend their week arguing about (supply, margins, Rubin timing) matters far less than the frame Jensen keeps reinforcing: compute equals revenue. Tokens are money. Once every hyperscaler CFO internalizes that, capital allocation debates become self-fulfilling. The Groq deal is worth watching closely. Jensen compared it to Mellanox. Mellanox became a $31 billion-per-year networking business in five years.

Then there is Microsoft. Azure grew 40%. AI annual recurring revenue hit $37 billion, growing at 123% year-over-year. That alone would be a top-300 S&P company. But the more important story is what CFO Amy Hood said about pricing: Microsoft is shifting from pure per-seat licensing to seat-plus-consumption across GitHub, Dynamics, and eventually across the whole 365 stack. When a CFO tells you a $600 billion-plus backlog is being repriced upward by usage, that is a fundamental change in the revenue ceiling.

Alphabet was arguably the most underappreciated of the five. Google Cloud (NASDAQ:GOOG) hit $20 billion in a single quarter. The AI solutions backlog nearly doubled to $462 billion. And then CFO Anat Ashkenazi mentioned, almost in passing, that 2027 capital expenditure will rise significantly above an already-raised $180-190 billion range. CFOs do not commit to that kind of spending without hard demand data. The TPU external sales announcement adds another layer: Alphabet is now in the enterprise AI hardware business at scale, which means that $462 billion backlog is probably underselling the opportunity.

Amazon's Andy Jassy dropped a figure that should have been front-page news: a $50 billion standalone chip valuation for Trainium. Amazon considers itself a top-three data center chip company. The market is not pricing that.

Tesla is the most complicated read. Elon Musk talked like a man building across a decade. $25 billion-plus in capital expenditure, two Optimus factories, the Terafab concept with SpaceX. CFO Vaibhav Taneja was plainly more cautious, flagging negative free cash flow for the rest of 2026 and real tariff exposure on energy margins. The tension between those two voices is genuine, and the Street has not figured out which one to weight.

The contrarian take across all five, considered together: the market keeps treating AI spend as a risk factor. These transcripts suggest it is becoming the closest thing to guaranteed ROI in corporate history. That gap between perception and reality is where the money gets made.

✦

Quick Takes

Three more stories worth knowing about

Azure at 40% Growth Means the Cloud Wars Are Over

Microsoft just posted Azure growth of 40% on a base that would make most tech companies dizzy. For context, this is not a small cloud catching up. This is the second-largest cloud on earth accelerating. The shift to seat-plus-consumption pricing is the part worth watching in 2026 and 2027. Once usage meters are attached to Microsoft 365 Copilot at scale, the revenue per seat stops being a fixed number. The installed base of 400 million-plus commercial users becomes an open-ended billing event. Amazon Web Services and Google Cloud are both growing fast, but Microsoft's distribution moat inside enterprise IT is genuinely different.

Tesla's Optimus Bet Is Bigger Than the Stock Price Reflects

Strip away the near-term noise (margin beats padded by one-time items, tariff drag on energy, negative free cash flow guidance through year-end) and what Tesla is actually building looks like nothing in the automotive sector. Two dedicated Optimus factories. A research chip fab. The Terafab concept with SpaceX for manufacturing at scale. Musk framed FSD as the core product and the vehicle as the delivery mechanism, which is either the most important reframing in the company's history or expensive storytelling. The 1.3 million paid FSD subscribers with falling churn gives that claim more weight than it had a year ago.

Google's TPU Move Changes the AI Hardware Math

Alphabet's decision to sell TPU access externally at enterprise scale is being treated as a footnote. It should not be. Google has spent over a decade and tens of billions of dollars building custom silicon optimized for its own AI workloads. Opening that to outside customers means the $462 billion Cloud backlog is sitting on top of a hardware revenue stream that is not in anyone's model yet. NVIDIA is still the dominant force in AI compute, but the competitive picture in 2027 and 2028 looks meaningfully different than 2024. Google, Amazon with Trainium, and Microsoft via Azure silicon partnerships are all moving in the same direction.

✦

Overlooked Stock

SKYT

SkyWater Technology, Inc.

SkyWater Technology (NASDAQ:SKYT) is the only pure-play semiconductor foundry operating entirely on U.S. soil, and it is sitting in the middle of one of the most politically durable spending trends of the decade: domestic chip manufacturing. SkyWater serves aerospace and defense, advanced computation, and bio-health customers, all sectors where supply chain sovereignty is not a preference, it is a requirement. As geopolitical pressure on Taiwan and Asia-based fab capacity keeps rising, the case for a homegrown alternative grows with it. The company is not trying to compete with TSMC on bleeding-edge nodes. It competes on specialized, rad-hard, and custom processes where foreign alternatives create unacceptable risk. That is a narrower market, but it is also a stickier one. Revenue has been growing, defense contract activity is expanding, and the CHIPS Act funding environment remains favorable. SKYT is small, thinly covered, and genuinely useful to customers who cannot afford to lose access to it. That combination does not stay cheap forever.

The Contrarian Take

The AI CapEx Panic Is Backwards — Underspending Is the Real Risk

Every few weeks a new research note warns that hyperscaler AI spending is a bubble waiting to pop. The argument is always the same: too much capital chasing too little return, history rhymes with the fiber glut of 2001, someone will blink first.

Here is the problem with that view: the customers are not blinking. Google Cloud's AI solutions backlog grew 800% year-over-year. AWS re-accelerated to 28% growth on a $150 billion run rate. Azure hit 40%. These are not companies spending ahead of demand. They are spending to keep up with it.

The real risk in 2026 is not overbuilding. It is the company that cuts capital expenditure in a moment of caution and loses three years of compounding to a competitor that did not. The executives running these businesses understand that. The analysts writing cautious notes about CapEx discipline are the ones who missed the last four years of this move.