Hook
Jane Street just dropped a 540% position increase into SanDisk. Let that sink in. The quant shop known for market-neutral precision isn't making a charitable bet on a random NAND flash maker. No. This is a signal. And most of the market is looking at the wrong end of the trade.
Here is the raw data. SanDisk, freshly spun off from Western Digital in February 2025, is trading at a PE somewhere in the 25-35x range. That is expensive for a memory company. Historically, these businesses trade at 15x when the cycle turns. But the 939 billion in long-term supply agreements with eight customers, including three major US cloud providers, changes the entire risk profile. This isn't just another cyclical storage name. This is a re-rating event.
Context
I have been tracking this space since the 2017 Parity multisig race, and I have seen pattern after pattern of institutional money moving before a narrative gets mainstream. Jane Street is not a retail-friendly shop. They trade on data and signal extraction, not vibes. So when they take a 540% position, I want to know what the market is missing.
SanDisk sits in a strange position. They are a pure-play NAND flash maker, having inherited the Flash Ventures fab partnership with Kioxia. They own the design, the fabs, and the sales channel. Their technology roadmap is currently at 218-layer BiCS8, with a 300-layer plus generation targeted for 2026-2027. Their QLC technology is best-in-class. But the real differentiator, the thing that Jane Street is likely betting on, is their High Bandwidth Flash (HBF) initiative. It's targeted at AI inference workloads, and it's positioned to compete in the same space as SK Hynix's HBM, but with a NAND foundation. And no one else has this in their portfolio.
Core
Let's break the financials down. The 939 billion in contracted revenue is not a number to ignore. That is multi-year visibility, something almost unheard of in the memory industry. It means that SanDisk's income is no longer exposed to the volatility of spot NAND prices in the same way. The contract is with the largest cloud service providers, who are essentially locking in supply for AI buildouts. The data center revenue grew 437% in the 2026 fiscal year. That's not a typo. That's the AI inference engine running full throttle.
But here's the part I keep coming back to in my analysis, and it's the core thesis. The industry is at the beginning of a storage upcycle, and the market is still pricing SanDisk like a memory company that will see a 20% drawdown. The current cycle is not being driven by smartphone demand or enterprise refresh cycles. It's being driven by AI inference, which is a structural growth engine, not a cyclical one. AI inference requires massive parameter storage and low-latency reads. That's exactly what SanDisk is building with HBF. It's a new market, not a share gain in an existing one.
Now, let's look at the broader landscape. SanDisk's capex intensity is projected to be 25-35% of revenue. That's heavy. The depreciation on new fabs will pressure gross margins by 2-4 percentage points. But the revenue growth from the long-term contracts is going to offset this. The operating cash flow is projected at $2-3 billion, and the free cash flow is positive, even with the capex load. The company has a ROIC above its WACC. That's value creation. The valuation is not cheap, but it's not irrational when you factor in the visibility.
Contrarian
The consensus narrative is that SanDisk is simply a derivative of the AI trade. The consensus is that if Nvidia stumbles, all storage names will be punished. That's too simplistic. The contrarian view is that the real opportunity is in the infrastructure that doesn't require the most advanced logic chips. SanDisk is a player in the NAND space, which uses DUV lithography, not EUV. They are immune to the most aggressive export controls. They have supply chain diversity via their Japanese and US fabs. Geopolitically, they are a safer AI play than most logic chip companies. This is a hedge, not a beta play.
Here is the more nuanced take, and this is where I want to be adversarial with my own thesis. The real risk isn't AI demand. The risk is the technical route. HBF is a new category. The market is going to compare it to HBM, and HBM is dominated by SK Hynix and Samsung. If HBF fails to gain traction, or if the client validation cycle takes too long, the $500 million to $1 billion in R&D spend might not get a return. The 20-25% probability of this happening is not zero. But the alternative is the potential of a $1-2 billion revenue stream in 2027-2028. The asymmetric payoff is still compelling. The stock is already up. The key is to watch the HBF sampling progress in 2026. If they hit the milestone, the current PE is a gift.
Takeaway
I'm tracking this position for the long haul. Jane Street's move is a classic "cheetah" sprint, but the underlying asset is a marathon. The next catalyst is the Q3 2025 earnings report, where I'll be looking at the data center revenue mix and the capex guidance. If the data center revenue growth holds, and the margin guidance is not weak, the market will have to re-rate this name as a growth story, not a cyclical one. The narrative of the market is shifting from the "chip wars" to the "memory wars." SanDisk is positioned to be a leader in the memory war. Stay alert. The next move is the next data point. — Root: The ESTP
— Cheetah
— Root: The ESTP