As we discussed earlier this month, it’s reasonable to think that some companies will be disrupted by AI. For some, AI will be existential. And AI will outright transform other companies — turning the old into new.
The latter is happening in the data storage business.
The earnings we heard from Western Digital and Sandisk this afternoon (we own both in our Billionaires Portfolio), suggest we may be witnessing the “Nvidia moment” for storage.
Remember, back in May of 2023, Nvidia shocked the world with a huge quarter, and then told us the data center demand was so steep, that the next quarter revenues would jump by more than 50% (from $7 billion to $11 billion). And they said the hyper-growth would continue for the foreseeable future.
So, May 2023 was the moment the world realized that AI was about to reinvent computing, and it would be powered by Nvidia’s GPUs. Similarly, January 2026 looks like the moment the world is realizing it needs endless data storage to feed the AI.
With that, today Sandisk delivered guidance that looked a lot like Nvidia’s famous Q2 2023 guide for a $7 to $11 billion quarterly revenue jump.
For the most recent quarter, Sandisk reported $3 billion in revenue (up 31% from the prior quarter) and then guided for $4.4-$4.8 billion next quarter(!). That’s a 50% sequential leap in just 90 days.
And margins are exploding. Gross margins were up 21 percentage points from the prior quarter, and they are guiding another 15 percentage points of margin expansion for next quarter.
What’s this all about?
Remember, as Jensen Huang said earlier this month at CES in Las Vegas, the data that AI produces is growing at a rate of 5x per year.
The AI models are running non-stop, creating oceans of data. Demand for data storage is “meaningfully” exceeding supply. And the data storage companies have been effectively sold out now for two quarters. This means pricing power – prices up, margins up.
And like Nvidia, Sandisk is now choosing its customers rather than chasing, and planning production against a committed backlog.
