Sandisk's Datacenter Breakout Shows NAND Has Stopped Behaving Like Commodity Flash

Sandisk's April 30, 2026 earnings release was the sort of quarter that forces a framework reset. Revenue reached $5.95 billion, up 97% sequentially. Gross margin hit 78.4%. GAAP net income was $3.615 billion. Free cash flow was $2.993 billion.

Those are not the numbers of a commodity supplier waiting for the next replacement cycle. They are the numbers of a company that is re-pricing its mix toward the highest-value parts of the storage stack, with datacenter now doing the heavy lifting.

The company said revenue outperformance was driven by a deliberate mix shift toward higher-value customers, with datacenter up 233% sequentially and pricing still improving. That is the key signal. Sandisk is no longer relying on a broad NAND recovery; it is leaning into customers whose demand is tied to AI infrastructure buildout.

The business model is also changing. Management said the quarter ended with three signed New Business Model agreements and that two more were signed in fiscal Q4. That sounds a lot less like spot-market NAND and a lot more like multi-year capacity planning with firm financial commitments.

This is what a transition from cycle repair to structural rerating looks like. Revenue, margin, and datacenter mix all stepped higher in sequence. The pattern matters more than any single quarter because it shows the business did not just spike once; it kept compounding at a faster rate.

The analogy I would use is not a one-time commodity rally. It is closer to a toll road getting busier, then getting repriced, then getting a bigger capital plan. If the traffic is real, the economics get better fast. If the traffic fades, the road still exists, but the premium disappears quickly.

The jump from $213 million to $1.467 billion in three quarters is the clearest evidence that Sandisk's mix is changing. The market usually gives memory companies credit only when pricing is tight. Here, the bigger point is that AI infrastructure demand is making the highest-value storage segments look structurally more important.

That matters because storage is often treated as an afterthought in AI. In practice, it is part of the control plane for data movement, latency, and total system economics. Once a customer designs around that constraint, the supplier can earn a very different margin profile.

My base case is that the market keeps rewarding Sandisk if datacenter remains the mix driver and the NBM structure keeps turning revenue into something closer to contracted capacity. If that happens, the stock deserves a much less skeptical multiple than a normal flash company.

The risk is equally simple: if AI buildout slows or mix normalizes, flash can revert to being a cyclical commodity very quickly. So the correct analogy is not a software annuity. It is a toll road with much better traffic than before, but still traffic-dependent.