NVDA, MSFT, GOOG, AMZN, TSLA all said the same thing

Five Earnings Calls, One Uncomfortable Truth for Everyone Sitting on Cash

Last week was the most concentrated display of corporate firepower in recent memory. Five of the biggest technology companies on earth reported earnings within days of each other, and the message across all five was basically identical: the AI spending cycle is not slowing down. It is speeding up.

Start with NVIDIA Corporation (NASDAQ:NVDA). Jensen Huang dropped the number ($62 billion in data center revenue, up 75% year over year) and then spent the rest of the call explaining why that number is almost beside the point. His framing was elegant and worth understanding: compute equals revenue, tokens are money, performance per watt is dollars per watt. Once every hyperscaler CFO fully accepts that logic, and most already have, the CapEx debate becomes circular. You spend more because spending more directly generates more. The Groq deal, which Jensen explicitly compared to the Mellanox acquisition, barely registered in headlines. Mellanox became a $31 billion per year networking empire within five years of being absorbed. Worth keeping in mind.

Microsoft Corporation (NASDAQ:MSFT) was arguably the cleanest call of the five. Azure at 40% growth, AI annual recurring revenue at $37 billion growing 123% year over year, and 20 million paid Copilot seats. But the number that matters most came from CFO Amy Hood, almost in passing: Microsoft is shifting from pure per-seat pricing to seat-plus-consumption across its entire product suite. When you reprice a $600 billion-plus revenue backlog upward via usage meters, that is not a product update. That is a fundamental change to the revenue ceiling of the entire company.

Alphabet Inc. (NASDAQ:GOOG) had a sleeper story that most analysts glossed over. Cloud backlog nearly doubled to $462 billion in a single quarter. Google is now selling TPUs externally at enterprise scale, which means the addressable market attached to that backlog is larger than anyone currently models. CFO Anat Ashkenazi said 2027 CapEx will significantly increase above an already-raised $180 to $190 billion range, and she framed it around tangible demand. Not optimism. Demand.

Amazon.com Inc. (NASDAQ:AMZN) had AWS re-accelerate to 28% growth, the fastest pace in 15 quarters on a $150 billion run rate. CEO Andy Jassy dropped a figure almost in passing: Amazon's chip division, Trainium, carries a standalone valuation of roughly $50 billion by his own assessment. Almost nobody is pricing that in.

Then there is Tesla Inc. (NASDAQ:TSLA), which was the most complicated call of the group. Elon Musk sounded like a man with a ten-year horizon and zero patience for quarterly thinking. CFO Vaibhav Taneja was flagging negative free cash flow for the rest of 2026, tariff exposure on energy, and one-time margin benefits. Both can be true simultaneously. The 1.3 million paid FSD subscribers, with falling churn and rising miles per customer, is real data that supports Tesla's reframing of itself as a software company that happens to manufacture vehicles.

Here is the uncomfortable part for anyone holding cash and waiting for a pullback to buy these names. The combined CapEx commitments from these five companies alone for the next 12 to 24 months likely exceeds $400 billion. That money funds NVIDIA's next revenue leg, Alphabet's TPU business, Amazon's chip ambitions, and Microsoft's consumption repricing. The spending is self-reinforcing. The only genuine bear case is a simultaneous ROI collapse across all hyperscalers at once. That is possible. It is not the base case.

✦

Quick Takes

Three more stories worth knowing about

Google's TPU Move Is Bigger Than the Cloud Numbers

When GOOG announced it will sell TPUs externally at enterprise scale, most coverage treated it as a footnote to the Cloud beat. That is a mistake. Google has spent over a decade building proprietary AI silicon that outperforms the market in specific inference workloads. Selling that hardware externally means the $462 billion Cloud backlog now has an attached hardware upsell attached to it. It also puts Google directly in NVDA's addressable market, not as a replacement but as a complement that captures spend Google previously left on the table.

Microsoft's OpenAI Deal Was Always About Cost, Not Friendship

The renegotiated Microsoft-OpenAI structure (royalty-free model IP through 2032 in exchange for OpenAI becoming a paying Azure customer) is the most financially significant partnership restructuring in tech this year. MSFT converted a famously expensive relationship into a long-duration cost advantage while simultaneously locking OpenAI's compute spend into Azure. Analysts covered it as a diplomacy story. It was a procurement story. The savings flow directly into Azure margins over the next six years, and the usage revenue from OpenAI is additive on top.

Amazon's Leo Satellite Play Is Being Assembled in Plain Sight

Andy Jassy's comment that the Amazon Leo satellite project, in combination with GlobalStar and Apple infrastructure, is 'reminiscent of AWS' in its capital intensity and long-run free cash flow profile got approximately zero traction in post-earnings coverage. That deserves a second look. AWS itself was invisible to most analysts for its first four years. If Jassy is drawing that specific comparison, in a public earnings call, with AMZN's track record behind him, dismissing it as executive boosterism seems like the riskier interpretation.

✦

Overlooked Stock

SKYT

SkyWater Technology, Inc.

While the market fixates on NVDA and TSMC (NYSE:TSM), SkyWater Technology (NASDAQ:SKYT) is doing something almost nobody else can: manufacturing semiconductors entirely on American soil, for American defense and aerospace customers who legally cannot use foreign-made chips.

SKYT is not a commodity foundry competing on cost with Taiwan. It is a specialty shop (MEMS, rad-hard chips, mixed-signal analog) that co-develops processes directly with customers. That model creates sticky, long-duration contracts rather than spot pricing exposure.

The CHIPS Act tailwind is real but secondary to the structural demand story. The U.S. Department of Defense cannot rely on Taiwan-based fabrication for its most sensitive programs. SkyWater is one of the very few companies positioned to fill that gap domestically. The stock trades at a fraction of the valuation applied to more glamorous semiconductor names, and the customer base is government-funded and largely recession-resistant. It is not fast growth. It is durable, defensible, and underpriced relative to its strategic importance.

The Contrarian Take

The Real TSLA Bull Case Has Nothing to Do With Cars

Most Tesla bulls talk about vehicle margins, Robotaxi timelines, or Optimus factory capacity. Most bears point to free cash flow, Musk distraction, and Chinese EV competition. Both camps are debating the wrong company.

The actual bull case, buried in Taneja's prepared remarks and hiding in the subscriber data, is this: Tesla has 1.3 million people paying a recurring monthly fee for software running on hardware they already own. Churn is falling. Miles driven per subscriber is rising. That is a SaaS business attached to a $300 billion installed base of physical assets.

No traditional auto company can replicate this. They do not have the data, the training infrastructure, or the vertical integration. If FSD genuinely reaches Level 4 reliability at scale (a real if, not a certainty), the licensing and subscription revenue potential dwarfs anything in Tesla's current income statement. The car business funds the research. The software is the product. That story is not priced in yet.