Every investor wishes they had anticipated this moment. When SanDisk’s stock was just separated from Western Digital, the market had already determined that it knew what SanDisk was: a cyclical memory maker with an uncertain future and a business model too dull to be trusted. At that point, the stock was trading at about $64 per share. After a full year, the price of SNDK shares reached $2,354 before returning to about $1,692. That represents one of the biggest single-year gains in semiconductor history, at about 2,440%.
It did not appear out of thin air. CEO David Goeckeler was clear about what SanDisk had become when it broke away from Western Digital and started trading on its own: it was no longer involved in legacy hard drive operations and was solely focused on NAND flash memory. According to reports, Goeckeler informed investors that he believed he had finally arrived at “the starting line of where the real value creation is going to happen.” It’s the kind of claim that seems practiced until the data begins to support it.
When the numbers were presented, they were difficult to dispute. According to SanDisk’s latest quarterly results, the company’s revenue increased by 372% from $1.9 billion to $8.97 billion. The gross margins were 84.6%. $6.9 billion was the net income. Additionally, the company set a goal of up to $10.8 billion for the next quarter. These are not the financial results of a business that happened to be in a fortunate quarter. This company reorganized, secured long-term contracts, and then, at the perfect moment, capitalized on a surge in demand for AI-driven storage.
The fact that SanDisk subtly shifted away from spot-market NAND transactions is what the majority of the larger market overlooked and is still underappreciated. Eight data-center clients, including three significant US hyperscale operators, have signed multi-year supply agreements with the company, totaling a minimum contract value of $93.9 billion. With floor pricing that supports gross margins of about 80%, these contracts cover over half of bit shipments through fiscal 2027 and about two-thirds in fiscal 2028. In the memory industry, that level of revenue visibility is uncommon. This is one of the reasons why some analysts now have a different perspective on the company.
This is not a hype story about AI. It is structural. Large amounts of fast storage, not just GPUs, are necessary for training and operating large language models. According to the Chief Product Officer of SanDisk, by 2030, the installed capacity of persistent KV cache storage alone could reach one zettabyte. The core of that demand is flash memory, and SanDisk and its joint venture partner Kioxia control roughly 33% of the global wafer production capacity. When data centers are growing at the current rate, that is a significant position to occupy.
However, there are some issues with the SNDK share price narrative. After momentarily failing to hold above $1,800 earlier in the week, shares fell back to about $1,692 on September 11. Senior executives, including the chief legal officer and chief technology officer, sold millions of dollars’ worth of shares under prearranged trading plans, according to regulatory filings. Insider selling close to historic highs has a way of making investors more cautious, especially when analyst price targets vary as widely as $1,400 from Wells Fargo to $2,450 from Lynx Research. However, these are scheduled transactions, not necessarily a sign of panic.
Additionally, there is the issue of competition. Lower-cost supply from Chinese manufacturers, especially YMTC and CXMT, may eventually put pressure on prices as they increase their production capacity. When you combine that with SanDisk and Kioxia’s joint commitment to invest over $31 billion in Japanese manufacturing facilities—roughly 60% of total spending in the nation over the previous 25 years—you have a company placing extremely high bets on the longevity of NAND demand. The investment appears prophetic if that demand continues. The math becomes more difficult if supply eventually surpasses it.
The fact that SanDisk’s next significant growth driver isn’t yet priced into the base case is one advantage. Co-developed with SK Hynix and designed for AI inference applications, High Bandwidth Flash technology is scheduled to launch in 2027 and go into mass production in 2028. Eight to sixteen times the memory capacity of existing solutions at comparable bandwidth is what HBF is intended to provide. Tenstorrent, Meta, and Alphabet have all expressed support. The financial targets management has set do not account for the potential upside if that technology is implemented as intended.
A few factors will determine where SNDK’s share price goes from $1,692: memory pricing remaining stable, AI infrastructure spending continuing, and the company’s capacity to fulfilll contracts that represent years’ worth of future revenue. After such a big run, the stock might need some time to settle. A memory company that has begun acting more like a software company—predictable revenue, high margins, long-term contracts—seems to be in the process of being valued by the market. Repricing is not a linear process. However, despite the recent retreat, the underlying narrative is still more intact, at least for the time being.
