NVIDIA's Groq bet could be the next Mellanox

Jensen Huang Is Building Something That Has No Comparable

There's a moment on every great earnings call when the company stops reporting results and starts describing a different kind of future. Jensen Huang hit that moment about twelve minutes in, and he never really came back.

NVIDIA's Q4 numbers are almost beside the point at this stage. Data center revenue at $62 billion, up 75% year-over-year. A Q1 guide implying nearly $300 billion in annualized revenue. Those figures are staggering, but they're also backward-looking. What matters is the argument Jensen is making, and whether it holds.

Here's the argument in plain English: compute is money. Every token generated is revenue for someone. Every watt of performance delivered is a dollar of value created. Once every hyperscaler CFO internalizes that framing (and they are internalizing it, fast), the debate about AI CapEx stops being a debate. It becomes circular. You spend more because the return is measurable. The return grows because you spent more.

That's not spin. It's structurally true, and it's why AWS, Azure, and Google Cloud are all guiding CapEx higher into 2027 despite every macro headwind you can name.

But two disclosures from this call deserve more attention than they're getting.

First, the $10 billion Anthropic investment. NVIDIA is no longer just the infrastructure layer for AI. It's now financially tied to the frontier model layer too. That's a different kind of moat. It means NVIDIA wins whether the compute goes up or the models get smarter. Both roads lead home.

Second, the Groq acquisition. Jensen compared it to the Mellanox acquisition in 2019. That comparison shouldn't be brushed off. Mellanox was a $7 billion deal that became a $31 billion per year networking empire within five years. If Groq does something similar for inference infrastructure, we're talking about another compounding engine bolted onto an already-compounding machine.

Colette Kress was measured throughout, flagging gaming supply constraints and acknowledging uncertainty around the Rubin ramp timeline. That's not a warning sign. That's an adult keeping the books honest while the founder bets the decade.

Now for the honest counterargument. NVIDIA trades at a premium that assumes everything goes right. Rubin needs to ramp cleanly. Groq needs to become Mellanox. Anthropic needs to matter at the model layer. Custom silicon from Google, Amazon, and Microsoft keeps losing. Regulatory scrutiny stays manageable. That's a long list of dependencies, and any one of them can disappoint.

The bear case isn't crazy. It's just losing, for now.

What NVIDIA has built is a de facto standard. The CUDA stack, the NVLink fabric, the software ecosystem. Standards are extraordinarily hard to dislodge once enterprise software is written on top of them. The switching costs are real and growing.

At this revenue velocity, with this capital return profile and this strategic optionality, the question isn't whether NVIDIA is expensive. It's whether the conventional tools for valuing semiconductor companies are even the right tools anymore.

✦

Quick Takes

Three more stories worth knowing about

Microsoft Just Made Its OpenAI Deal Look Like a Bargain

The renegotiated OpenAI partnership is the most underreported story from this earnings season. Microsoft gave up some equity upside. In exchange, it got royalty-free model IP through 2032 and turned OpenAI into a paying Azure customer. That's not a diplomatic compromise. It's a long-duration cost advantage disguised as a friendship. Azure at 40% growth and an AI ARR of $37 billion growing 123% year-over-year are the headline numbers, but the business model shift (per-seat plus consumption metering across GitHub, Dynamics, and eventually M365) is what reprices the entire installed base upward.

Google Cloud Is No Longer Playing Catch-Up

AWS and Azure got most of the cloud attention this cycle, but Alphabet's numbers demand a second look. Cloud backlog nearly doubled to $462 billion in a single quarter. AI solutions revenue grew 800% year-over-year. The TPU external sales announcement (Google is now selling its custom AI chips to enterprise customers) means the addressable market behind that backlog is larger than any current model assumes. CFO Anat Ashkenazi flagging a 'significant increase' in 2027 CapEx above an already-raised $180-190 billion range, based on 'tangible demand signals,' is the kind of statement that deserves to be taken literally.

Amazon's Chip Business Is Worth $50B and Almost Nobody Prices It In

Andy Jassy dropped a number almost casually on the Q1 call: Trainium, Amazon's custom AI chip, is worth roughly $50 billion as a standalone business by their own assessment. That figure sits inside a stock the market still largely values as a retail and cloud company. AWS re-accelerated to 28% growth (fastest in 15 quarters) on a $150 billion run rate. The OpenAI partnership via Bedrock Managed Agents creates a stickiness play with no direct equivalent anywhere. And Amazon Leo plus Globalstar plus Apple is a trillion-dollar adjacency being assembled in plain sight. Jassy compared it to early AWS. He's done that comparison before and been right.

✦

Overlooked Stock

SKYT

SkyWater Technology, Inc.

While the market fawns over TSMC and Samsung, SkyWater Technology is doing something no other US-soil foundry can: making chips on American soil for customers who legally cannot use foreign fabs. Defense contractors, aerospace engineers, and government programs that require ITAR-compliant, US-manufactured semiconductors have almost nowhere else to go. SkyWater is it.

The company operates out of Bloomington, Minnesota, running a pure-play foundry model focused on analog, mixed-signal, MEMS, and radiation-hardened chips (the unglamorous but irreplaceable components that go into satellites, weapons systems, and medical devices). CHIPS Act funding and the broader push toward domestic semiconductor sovereignty are tailwinds that are structural and policy-driven, not cyclical.

This isn't a speculative moonshot. It's a niche foundry with a defensible moat, a captive customer base, and a geopolitical backdrop that keeps improving. The stock is thinly followed and deeply misunderstood. That's usually where the opportunity lives.

The Contrarian Take

Tesla's Car Business Is Already Irrelevant to the Bull Case

The Tesla bears are still fighting last year's war, arguing about margins, deliveries, and whether Elon has lost the plot. That debate misses what the Q1 call actually said.

CFO Vaibhav Taneja confirmed 1.3 million paid FSD subscribers with falling churn and rising miles per customer. Then he said the quiet part out loud: 'FSD as the product, vehicle as the delivery mechanism.' That's a software company describing its distribution strategy, not an automaker defending a price cut.

Negative free cash flow through the rest of 2026 is real and the tariff exposure on energy hardware is real. The Optimus and Robotaxi timelines carry genuine execution risk. But if Tesla converts even a fraction of its fleet into a recurring software revenue base, the auto margin debate becomes a distraction.

The Street is pricing Tesla as a car company with optionality. It should be pricing it as a software company that happens to own 7 million rolling data collection devices. Those are very different valuations.