Nvidia's Earnings Setup
Nvidia (NVDA) has become the center of the AI trade and arguably the most important company in that trade. Ahead of its earnings report, one worry stands out: in six of the last eight earnings reports, the stock traded down the day after, even though these were beat-and-raise quarters. In seven or eight of the last eight, it was still down a month later. The report also comes after a 7-day decline, the longest losing streak since September 2022.
My prediction: the company beats and raises. Why a stock falls after a beat and raise is hard to explain, and someone else will have to account for that pattern. A normal reason for a post-report drop is a big run-up before the print - a "sell the news" event where a lot is already priced in. That does not fit here, since the stock fell into the report, which makes this drop feel odd.
Demand Is Real and Instant
The fundamentals are strong. Hyperscalers have posted $800 billion in projected forward investment. Nvidia is projecting price increases, which means it is holding its margins. Demand is strong across the whole sector.
Talking with senior leaders at AWS, Google, and elsewhere, the message is the same: rampant demand. A Neo cloud provider CEO cannot get GPUs fast enough; every batch sells out on arrival. Every GPU that comes online is consumed as soon as it is available. This is not a build-into-dark-demand case like broadband in the 1990s. Everything that comes on board is monetized right away and put to work.
A fresh sign of demand: a Perplexity collaboration announced that day to bring AI to the desktop.
From Chip Maker to Full-Stack Player
Nvidia (NVDA) looks less and less like just a chip company. It now offers financing, CPUs, networking, and software layers. It has taken a leadership position in AI networking, is bringing its own models to bear with Nemotron, and is working with Perplexity on the desktop. It sits up and down the stack, and CUDA remains a huge moat.
Being full-stack makes it harder for customers to move away from the Nvidia ecosystem, but customers will still make choices. Intel (INTC) is in the picture, and ARM (ARM) announced work this week on inference on the CPU layer.
Will Nvidia keep 90%-plus share in GPUs? That share will erode over time. But this is a rising-tide scenario - classic Jevons paradox, where more demand for compute drives more demand across the board. Losing some share is not a negative drag on total platform demand.
Financing and Depreciation Questions
On whether Jensen Huang will address depreciation and circular financing: he will not raise it in his main remarks, but the equities analysts will not let him off the hook, so expect questions.
On the $500 billion Nvidia is putting in place to help finance the buildout: this is like Ford (F) financing cars, not a circular economy. GPUs are becoming collateralized assets. Some Neo clouds, such as Cumulus AI under CEO Mike Maniscalo, are finding new ways to finance and bring on systems.
This is a huge buildout with obligations of up to $800 billion. It is very early - only about 3 years in - and this looks like a 10 to 15 year trend, so the focus on how to finance it will keep growing.
Blackwell vs. Vera Rubin
Many previews stress Vera Rubin and its scaling, and the market seems to be looking past Blackwell, wanting proof the growth cycle runs beyond 2027.
The real story is the speed of innovation. Nvidia brings new systems to market on a yearly cadence, which is unheard of against other GPU and semiconductor makers. But do not read Vera Rubin's arrival as Blackwell being dead. H200s and B300s - the platforms before Blackwell - still see rampant demand. Frontier labs doing training will want the cutting-edge systems, but a lot of inference still needs to be built out across the Neo cloud space, and that will use products further down the cycle.
Blackwell demand will stay strong. Every GPU Nvidia can make is gobbled up by Neo clouds, hyperscalers, and, increasingly, enterprises. Enterprise demand is behind the explosive quarters and earnings at Dell (DELL), HPE (HPE), Lenovo, and Super Micro (SMCI).
The Biggest Risks
China remains a wild card with the H200s. For many quarters, China has been treated as the icing on the cake, with everything priced in "ex-China."
Memory prices keep climbing, and Jensen Huang has been explicit about this. A 15% memory price increase is coming, which helps protect margins, but the question is how enterprises and hyperscalers absorb it.
Another risk on the edges: reports in the prior 48 hours that some consumer-type Nvidia GPUs turned up in Russian drones, which will draw scrutiny.
These are noise around the edges. The core fundamentals - hyperscaler growth, Neo cloud growth, and enterprise adoption - are all strong, and they set the overall trajectory.
