SK Hynix Hits Record Profits: Is Samsung Still a Buy in the AI Memory Race?

The headline numbers were extraordinary. Revenue rose to 52.58 trillion won, a 60 percent quarter-on-quarter increase 198% on a year-over-year basis. Operating profit jumped to 37.61 trillion won, up 96% from the previous quarter and 405% compared to the same period last year. That was a 72% operating margin, a rarefied level in high-CapEx memory manufacturing. Management said the increase was driven by higher prices for DRAM and NAND as well as a more favorable mix of high-value solutions, such as HBM for AI data centers and large capacity server DRAM and enterprise SSDs.

The larger backdrop is one of a memory market adjusting to AI server investments driving memory into a structural shortfall. Demand for the data center has pushed prices higher sequentially for both DRAM and NAND, as industry checks suggest tight supply. SK Hynix is responding with a rise in capital spending for long-term capacity—related to the M15X line, the Yongin cluster infrastructure, and EUV tools. The company has 54.3 trillion won in cash and equivalents and 35 trillion won in net cash at the end of the quarter, which affords it the ability to expand while keeping a solid balance sheet. For HBM in particular, SK Hynix is strengthening its cooperation with its customers for future nodes such as HBM4, and maintaining the leadership in the most supply-constrained segment.

Samsung also highlighted just how this upcycle is different, with the preliminary numbers for the first quarter already in. Revenue was about 133 trillion won, up roughly 68% year over year, and operating profit soared to 57.2 trillion won, which is about 7.5x the prior year and over 50% above consensus expectations. This would appear to support what was reported: HBM and other data center memory products are contributing to a supply crunch. The industry is seeing longer-term supply contracts, which are unheard of in memory, allowing suppliers to have better earnings visibility and reduced risk of a sudden contraction.

In comparison, SK Hynix had an eye-popping margin due to a product mix weighted more heavily toward HBM and premium DRAM. Samsung’s beat was larger in absolute profit and indicates it is also benefiting from the AI memory surge despite a more diverse product portfolio. That both sets of results serve up the same reality—the prices are with the memory producers and the constrained supply is probably not going away anytime soon given wafer constraints and HBM’s somewhat higher wafer intensity versus traditional DRAM.

Are memory stocks on earnings the better thing to look at in a boom? Because earnings are so volatile through cycles, the P/E can appear to be very cheap around the top end of the cycle and very expensive near the bottom. For this reason, a lot of investors track Price-to-Book with memory. By that measure, Samsung appears high relative to its own history. In a normal cycle, a higher P/B might be a warning sign. However, this cycle is not normal. HBM demand, tight wafer supply, and longer-term contracts support the case for profits remaining robust for a number of quarters. In that context, a premium to book value can be sensible if it is on the back of sustained high return on capital during a multi-year AI build-out.

Cyclicality continues to be the biggest danger. SK Hynix is ramping up capital expenditure, and its rivals are indicating larger budgets as well. As is ordinary, this is what softens hot markets: over time, more supply. The other two risks mentioned are productivity advances enabled by new software and model methodologies. Alphabet Inc. (GOOGL) (GOOG) has improved some of them, such as memory optimization with TurboQuant. Though they may moderate unit intensity over time, the current wave of AI is still bottlenecked on physical memory availability, and historical trends indicate efficiency gains tend to expand use rather than reduce demand outright. The more likely effect is that the next downward cycle could be more exaggerated on the upside when supply catches up. A third danger is reliance on data center security. If non-data-center segments such as PCs and smartphones stumble under supply constraints, the entire mix can become more heavily weighted toward AI server demand.

Samsung’s quarter is evidence of strong earnings recovery and cash generation prowess even in a tight market. The company is riding AI-driven pricing, a greater share of premium memory, and longer-duration contracts that smooth near-term volatility. The magnitude of its operating profit beat vs. consensus suggests analysts were blindsided as the AI memory prices surged. This lends support to the notion that Samsung will be able to provide for the required HBM and advanced DRAM investment while retaining balance-sheet strength. The business, in other words, is set to rampage through the up-cycle and prepare for the next cycle.

There is now a direct implication for SK Hynix’s stock: 72% operating margins and record revenue/profit show that AI-driven memory demand is creating real earnings power. For what's next, watch capacity decisions and product leadership in HBM to see how long premium pricing can last.

The case for Samsung stock could be made on three points. First, the same structural tightness that is driving SK Hynix’s margin expansion is also driving a profit recovery and improving visibility for Samsung. Second, Price-to-Book seems crowded, but the market may be rightly pricing in a multi-quarter run of high returns driven by AI memory scarcity. Third, although software inefficiency and capacity additions are real risk factors, they probably won’t end the cycle before more supply comes to market and demand growth normalizes, which traditionally is a process that takes some time.

Those looking for a U.S.-listed peer exposure could look to Micron Technology, Inc. (MU) – it too covers DRAM, NAND, HBM, and data center SSDs and is flagging similar tight supply environments going into 2026. There are also broader options players like Roundhill Memory ETF (DRAM), but when you have a concentrated memory exposure, it just makes position sizing that much more important with cyclicality. For a sense of where adjacent ecosystems tied to AI infrastructure are also benefiting, NVIDIA Corporation (NVDA) or Taiwan Semiconductor Manufacturing Company Limited (TSM) still serve as decent gauges, though their cycles and risk profiles are less pure-play memory.

On the whole, yes—but you need to understand memory cyclicality and size your position accordingly. Samsung’s Q1 earnings suggest the profit runway is growing due to HBM-driven shortages, a richer product mix, and long-term contracts smoothing the near-term path. The Price-to-Book is not low, but in these times an above-average multiple is warranted due to extraordinary profitability and limited supply. The key watch items are CapEx trajectories industrywide, signs of relief in HBM tightness, and whether efficiency gains from AI software start capping unit intensity by 2026.

In short, SK Hynix’s stellar quarter is proof that the upcycle is real across the premium memory landscape. The investment case for Samsung in 2026 is predicated on that same tightness. The potential is enticing, but it is still a cyclical play—treat it like one, not like a sure thing.