Maximilian Heitsch @MrMagoon: "The thing that makes $SNDK..."

Germany

The thing that makes $SNDK (Sandisk Corp/DE) so interesting right now… is that the market still doesn’t fully believe the story. And honestly? That skepticism is understandable. Memory stocks have traumatized investors for decades. Every cycle looked the same: prices up everyone expands capacity oversupply arrives margins collapse repeat. That’s why the current valuation looks so absurd. Depending on which estimates you use, $SNDK’s PEG ratio is somewhere around 0.2–0.4. Which sounds completely broken when you look at the numbers: Q3 revenue: +251% YoY Q4 EPS guide: $30–33 gross margins near ~78% Normally, stocks with numbers like that don’t trade around ~10x forward earnings. The market is basically saying: “Cool. But show me this still works in 2027.” And THAT is the real debate. Not whether the current numbers are strong. The real question is: Has AI permanently changed the NAND cycle? Because the business itself is evolving. This isn’t just smartphones and laptops anymore. This is: AI training AI inference vector databases RAG systems huge context windows massive enterprise storage demand And hyperscalers are now securing capacity years ahead. That part is critical. $SNDK already signed contracts covering at least $42B in future revenue. Large parts of 2027 production are already locked in. That’s a completely different dynamic versus the old spot-market chaos. But the really interesting part starts one layer deeper: HBF. High Bandwidth Flash. Basically: stacked NAND designed for AI workloads. Think of it as a new memory layer between: HBM and traditional SSD storage. The idea is simple: HBM is insanely fast… but extremely expensive and capacity-limited. AI systems increasingly need something in between: more capacity lower cost still very high bandwidth That’s exactly what HBF tries to solve. And if this actually works at scale… the entire valuation framework changes. Because suddenly $SNDK isn’t just selling commodity NAND anymore. It starts becoming part of AI infrastructure itself. That’s a completely different category. Now obviously: this is still early. There are still huge open questions: real-world performance thermal limits GPU integration software ecosystem actual hyperscaler adoption So this is not some guaranteed outcome. But the setup is fascinating. Especially because the market still treats NAND largely like an old cyclical commodity business. Meanwhile the underlying demand profile is changing dramatically. That’s why I keep talking about bottlenecks. Because every phase of this AI cycle creates a new one. First compute. Then memory. Then photonics. Now storage architecture itself. And the really crazy part? This entire AI infrastructure buildout still looks early. That’s why I’m still long $SNDK alongside names like: $MU (Micron Technology, Inc.) $WDC (Western Digital Corporation) $NVDA (NVIDIA Corporation) $AVGO (Broadcom Inc) $MRVL (Marvell Technology Group Ltd) $AAOI (Applied Optoelectronics Inc) The thesis may fail eventually. But right now the data still points in one direction: AI storage demand is accelerating faster than supply. And until that changes… the cycle stays intact. — I’m putting together a small private group of ~50 high-quality eToro investors. If you want in, visit my X account (link on eToro profile). I posted a link to the waitlist there. Once we reach 50 waitlist entries, it will be closed and the Telegram group will be opened. This is not investment advice, for education purposes only | Capital at risk | Past performance does not guarantee future results

Not investment advice. The author may have financial interests in the mentioned instruments.