Samsung’s memory business spent much of the past two years in a valley. Prices for DRAM and NAND collapsed after a pandemic-era buying binge, inventories piled up, and the company that had long printed money from selling memory chips suddenly found itself reporting operating profits that looked nothing like its peak. That slump became the defining test for Lee Jae-yong, who formally took control of the conglomerate four years ago and has spent the time since trying to convince investors that Samsung can convert its manufacturing muscle into a position at the center of the AI economy.
The turnaround narrative now resting on Lee’s shoulders is built around high-bandwidth memory, the specialized stacked DRAM that gets packaged alongside AI accelerators from Nvidia and others. HBM has become the most contested piece of silicon in the industry because every large language model, every training cluster, and every inference farm needs enormous amounts of it. Samsung was late to qualify its HBM products with key customers, a misstep that let a smaller rival seize the early lead in a market Samsung once assumed it would dominate by default. Lee’s job has been to close that gap without losing ground in the commodity memory business that still funds much of the company’s research.
What makes the current moment interesting is how much of Samsung’s recovery depends on execution rather than luck. The AI boom has created demand that is genuinely difficult to satisfy, and HBM yields are notoriously hard to master. Companies that get the stacking, thermal management, and packaging right can charge premium prices. Companies that stumble watch their would-be customers sign multi-year contracts elsewhere. Lee has pushed the organization to treat that reality seriously, reorganizing reporting lines and emphasizing speed over the cautious consensus-building that has sometimes slowed Samsung’s response to market shifts.
There is a broader pattern here that goes beyond one company. The chip industry’s center of gravity has moved decisively toward AI infrastructure, and the firms that supply memory, packaging, and foundry capacity are being revalued by investors who see them as essential plumbing for the compute buildout. Lee’s challenge is that Samsung competes on several fronts at once: memory against SK Hynix and Micron, foundry against TSMC, and consumer devices against everyone. Spreading resources across those fronts is both Samsung’s greatest strength and its most persistent vulnerability, because a company that does many things well can struggle to be the undisputed best at any single one.
The strategic logic Lee appears to be pursuing is integration. Samsung is one of the few companies that can design logic, manufacture memory, run a foundry, and package chips under one roof. If AI systems increasingly require custom silicon and tight co-design between memory and compute, that vertical reach becomes an advantage rather than a distraction. The pitch to customers is essentially that Samsung can be a one-stop partner for the increasingly complex stack that AI hardware demands. Whether customers actually want that level of bundling is an open question, and the answer likely varies by buyer.
Skeptics have fair points. Samsung’s foundry business has struggled to win marquee customers away from TSMC, and its yield challenges at advanced nodes have been widely discussed in the industry. Its consumer electronics divisions face mature, slow-growing markets. And the memory business remains cyclical by nature, which means any AI-driven upswing will eventually meet a correction. Lee’s tenure will be judged less on whether he catches the current wave and more on whether he builds enough durable advantage in AI-era components to soften the next downturn. That is a harder test than simply shipping more HBM.
The geopolitical backdrop adds another layer. Export controls, subsidy programs, and the push to diversify manufacturing away from concentrated geographic regions are reshaping where chips get made. Samsung has responded with fab investments in the United States and elsewhere, decisions that carry long timelines and heavy capital costs. Lee has to balance those commitments against shareholder pressure for returns and the operational discipline needed to keep existing plants competitive. It is the kind of multi-front capital allocation problem that few executives face at this scale.
Four years is a short window in the life of a conglomerate that measures its history in decades, but it is enough time to see the direction of travel. Lee Jae-yong has moved Samsung toward AI infrastructure at exactly the moment the industry’s economics demand it, and he has done so while managing the fallout from a memory slump that humbled the company. The next phase will reveal whether that repositioning is genuine strategic transformation or a well-timed ride on a boom. For now, the market is giving Samsung credit for the former, and Lee’s task is to keep earning it.