The alert went out before the candle closed. — On March 10, 2025, Sandisk’s first trading day as a standalone stock, I watched the tape from my desk in Dubai. The stock ripped 15% in the first hour. My phone exploded: “Is this the next NVIDIA?” I laughed. I’ve been here before. In 2017, during the Telegram sprint, I spotted a vulnerability in an ERC20 token’s minting function and rushed to publish a breaking news alert. That alert went out before the candle closed. The market reacted instantly, but the underlying code was flawed. The same dynamic is playing out with Sandisk. The market is reacting to the narrative, but the underlying code—the NAND cycle—is still the same. The noise fades, but the pattern remembers.
Context: Why Now? Let’s rewind. The NAND market has been a textbook cyclical industry since its inception. Every two to three years, a boom in demand from smartphones, then cloud, then SSDs leads to overinvestment, oversupply, and a brutal price collapse. The 2023-2024 downturn was one of the worst—operating losses across the board. Western Digital’s storage bled, forcing the spin-off of Sandisk as a pure-play NAND company. The logic: by separating from the HDD business, Sandisk can focus on the high-growth AI-driven enterprise SSD market. And indeed, the numbers look promising. AI servers require massive storage—each training run generates petabytes of data, and inference servers need fast access to model weights. Sandisk’s 218-layer BiCS8 NAND, its QLC enterprise SSDs, and its partnership with Kioxia (which manufactures the NAND) position it as a key player. The market is pricing in a new era where NAND demand grows at 10-15% annually, driven by AI, and the cycle is tamed. But I’ve lived this pattern before. I’ve been tracking NAND cycles since 2017, when I was a junior cybersecurity analyst. Back then, I monitored Telegram channels for ICOs, but also kept an eye on semiconductors because they correlated with mining hardware. The 2017 NAND boom was driven by smartphones. Then came the 2018 crash. In 2020, during the DeFi Summer, I was livestreaming on Twitch, analyzing Uniswap pools. I noticed NAND prices were bottoming. I called it on stream: “NAND is about to cycle up.” It did, driven by remote work and cloud. Now, in 2025, the narrative is AI. But the pattern is the same: a new use case emerges, demand spikes, prices rise, manufacturers add capacity, and then prices collapse. The question is whether AI is different.
Core: The Data and the Impact From my vantage point as a real-time trading signal strategist, I’ve been tracking NAND contract prices since the 2020 upcycle. The current trend: Q1 2025 saw NAND contract prices rise 5-10% quarter-over-quarter, with enterprise SSDs seeing even larger jumps (TrendForce data). That’s a typical recovery pattern. The inventory cycle is swinging from destocking to restocking. The key question: is this a one-time re-stocking or a sustained demand shift?
Let’s dig into the AI inference demand. A single inference server with 8 GPUs might need 30TB of SSD for model weights and context. If we assume 10 million such servers by 2027, that’s 300 exabytes of NAND demand. But that’s a big if. I’ve seen similar projections in crypto: “Layer2 will scale Ethereum to a million TPS”—then the data showed otherwise. From static streams to living liquidity—the NAND demand from AI is real, but the growth rate may be overhyped. The real story is the supply side. NAND manufacturers have a history of forgetting their pain. In 2023-2024, they lost billions. Now, with prices rising, the temptation is to ramp up capacity. I recall a conversation with a Micron supplier at a Dubai conference who said, “We always say we’ll be disciplined, but when the market is hot, we chase it.” That’s the psychological pattern. I’ve seen it in crypto mining: after the 2021 crash, miners promised hash rate caps, but within a year, the hash rate hit new highs. The same applies to NAND.
Sandisk’s specific position is interesting. Its 218-layer BiCS8 is competitive with Samsung and SK Hynix’s 300-layer efforts. Its QLC enterprise SSDs are targeting read-intensive AI workloads. But the secret sauce is the controller firmware and LDPC error correction. That’s where the real moat lies. However, Sandisk’s dependency on Kioxia for manufacturing is a single point of failure. During the 2022 crash, I hosted a networking dinner for crypto founders in Dubai. One guest was a former Kioxia engineer. He told me, “The joint venture is a marriage of convenience, not love.” The two companies compete in the same enterprise SSD market. That tension is a hidden risk. If Kioxia faces financial trouble or a natural disaster (Japan is prone to earthquakes), Sandisk’s supply chain is severed. The market is ignoring this vulnerability. We didn’t just watch the chart, we lived it—I’ve seen similar dependencies wreck projects in DeFi.
Contrarian: The Unreported Angle Here’s the angle everyone is missing. The narrative that “AI inference changes NAND’s cyclicality” is a manufactured story, just like the “liquidity fragmentation” narrative in DeFi that VCs used to push new cross-chain bridges. The reality is that AI inference may not consume as much NAND as the market hopes. Model compression techniques—quantization, pruning, distillation—are reducing the storage footprint of AI models. A 70B parameter model can be shrunk to 4-bit precision, cutting storage requirements by 75%. And as inference moves to edge devices, the demand for high-end enterprise SSDs may plateau. Shiny objects distract, but dry powder preserves. I’ve seen this in the NFT space: I identified a rug-pull project by checking the contract code. Similarly, I’m checking the “code” of the NAND cycle: the capex commitments. If Sandisk announces a major fab expansion, that’s a red flag. It would signal they’re drinking the Kool-Aid. The pattern remembers: every time a new use case was supposed to break the cycle, it didn’t. The cloud boom of 2017-2018 ended in a crash. The remote work boom of 2020-2021 ended in a crash. The AI boom will likely follow the same path, just with a longer duration.
Geopolitical risk is another blind spot. If US-China tensions escalate, Sandisk could lose access to the Chinese market, which is a significant consumer of NAND. The Biden administration’s export controls are expanding. NAND isn’t directly targeted yet, but enterprise SSDs could be. I’ve seen this in the crypto mining industry: when China banned mining, the entire supply chain shifted. The same could happen to NAND.
Takeaway: The Next Watch The next watch is Sandisk’s Q2 2025 earnings and their capex guidance. If they are cautious and maintain supply discipline, I’ll be bullish on the stock. If they go all-in on expansion, I’ll short it. The pattern remembers. And as I always say, trust the code, verify the art, ignore the hype. The code is the on-chain data: NAND prices, inventory levels, capex guidance. The art is the narrative. The hype is the AI story. Right now, the hype is loud. But I’ve been burned by too many “narratives” in crypto to take this one at face value. Keep your dry powder ready. The real signal will come when the noise fades.
We didn’t just watch the chart, we lived it. I’ve lived through multiple cycles, and this one feels different only because the hype is louder. But the fundamentals are the same. The noise fades, but the pattern remembers. The alert went out before the candle closed. Now, I’m watching the next move.