Samsung in the Shadow of Record Profit: Why the Shares Fall Even at the Peak of the AI Memory Cycle

Samsung in the Shadow of Record Profit: Why the Shares Fall Even at the Peak of the AI Memory Cycle In late June, Samsung Electronics shares touched an all-time high. Four months later, as the company guided toward a fourth straight quarter of record operating profit, those same shares sat more than 25 percent below that peak. The income statement is racing upward; the market is moving down. That is the central riddle here. Anyone reading only the headline might see a pure success story, but a share price never waits for past success; it pays for the future. When confidence in that future is missing, even a record becomes a warning. I have spent decades writing about markets and the economics of sport, and the same pattern keeps repeating: big numbers do not always tell a big story. Samsung's latest guidance is exactly one of those moments. Samsung said its operating profit for the July-to-September quarter of 2026 will come in at roughly 107.4 trillion won, about $80.17 billion. Revenue is around 195 trillion won, up 127 percent year on year. It is the fourth consecutive quarter in which the company has posted record operating profit. The LSEG SmartEstimate, the analyst consensus, stood at 106.1 trillion won. The company beat it slightly, not spectacularly. That is the first subtle signal. A record that only just edges past expectations is no longer explosive news. The market had already priced in an excellent result, so the good news did not lift the price; it fell instead. The interesting part is that the market's question is no longer "how big is the profit?" It is "how long does the profit last?" Kim Seok-hwan, an analyst at Mirae Asset Securities, put his finger exactly there: investors are now thinking about durability, not about the backward-looking record. That is a classic sign of the late stage of a cycle. When the numbers are highest, the market is already looking at the next turn. The key point is that Samsung's problem is not a lack of profit but the concentration of profit. Look inside the record numbers and most of the gain comes from the memory division. The company itself says the bulk of the earnings improvement came from memory. At the same time, the mobile division lost more than $1 billion, and the foundry business is expected to be loss-making with low utilisation. One division's spectacular rise is covering the persistent losses of two others. That covering is the most dangerous part. As long as memory prices keep rising, the weakness in mobile and foundry stays hidden. The day memory prices stop rising, those deficits will no longer be concealed; they will move to the front. And the pace of memory price growth is already slowing. According to TrendForce, DRAM contract-price growth was around 60 percent in the second quarter and may fall to 10 to 15 percent in the fourth. Operating-profit growth is forecast to slow from 20 percent quarter on quarter in the third quarter to 8.2 percent in the fourth. That is the real signal: record numbers, but a decelerating rate. The stock market always looks at the rate of change, not the level. A company growing profit at 20 percent that slows to 8 percent does not receive congratulations; it receives doubt. There is another layer to that doubt: currency. A large share of Samsung's revenue is dollar-denominated but reported in won. When the won strengthens, the won value of dollar sales falls. Because of this currency drag, analysts have trimmed their forecasts. It is not a dramatic problem, but it dulls the shine of the record. So where does Samsung stand on the competitive pitch? The most valuable product of the AI era is HBM, high-bandwidth memory. There the leader is SK Hynix. Samsung lags but is closing fast: bit shipments rose roughly 50 percent quarter on quarter, narrowing the gap with the leader. That is a positive. In conventional DRAM and NAND, Samsung is strong and a beneficiary of the price surge. The problem is foundry, where TSMC is the clear leader and Samsung trails in loss. Add rising competition from China, which could pressure margins over the medium term. Samsung, SK Hynix and Micron, the memory oligopoly, are all enjoying record margins because of a temporary imbalance between supply and demand. AI-infrastructure demand may sustain that imbalance until 2028, according to forecasts. But that is a forecast, not a certainty. This is where the value chain matters. The chain is not simple. Upstream is a memory supply shortage; midstream is record margins for Samsung, Hynix and Micron; downstream is AI-infrastructure demand. Together it looks bright. But the same demand raises component costs downstream, in consumer electronics and smartphones. Memory price inflation benefits the upstream oligopoly while lifting input costs for Samsung's own device divisions. It is an internal contradiction: the company profits with one hand and widens losses with the other. This cost pressure is one reason behind the mobile division's loss of more than $1 billion. Now notice the gap between the media story and market behaviour. Headlines say "world record," "$80 billion," the language of celebration. Market behaviour says caution: the share is flat to down, more than 25 percent below the June high. That divergence between media heat and price action is the strongest hint: the market is pricing a cycle peak, not current results. From years of watching markets, I have learned that when headlines and prices stop walking together, the price is usually right. Prices hold the expectations of the next three to six months; headlines hold the accounts of the past three. This is where I hesitate. Every strong claim should have a weak side, or it is propaganda rather than analysis. So how could I be wrong? First, if Samsung's HBM progress is faster than expected, if the 50 percent quarterly bit-shipment growth holds and the gap to the leader narrows further, memory profit could stay strong for several more quarters. The cycle peak would move further out. Second, if demand for advanced nodes returns to foundry, Samsung's losses there could shrink and the profit base could widen beyond memory. Third, if the detailed release on October 29 announces a shareholder-return policy, a buyback or a higher dividend, market sentiment could turn quickly. Analysts like Kim Seok-hwan are watching exactly that. Fourth, if AI capex runs as forecast through 2028, the demand base for memory prices will not collapse, and my claim that the peak is near would prove premature. Still, my core caution stands: the decoupling of record profit from a falling share price is itself a message. The market is pricing forward risk. So what comes next? I will give a dated prediction so it can be checked. If the detailed release on October 29, 2026 shows mobile and foundry losses larger than expected, further pressure on the share is likely. This prediction expires with the fourth-quarter results, around early January 2027. I will abandon the claim if two conditions are met: DRAM and NAND contract-price growth breaks back above the 10 to 15 percent range, and Samsung matches the leader's bit shipments in HBM. If either happens, the cycle story changes. The most important question now is this: is Samsung's story the story of one company, or a test of the entire AI era? If memory prices depend on the pace of AI investment, then Samsung's falling share is not just one company's worry. It may be a question about the biggest bet in the technology market: has the heat of the AI boom already reached its own peak? We will know more by October 29, and then by the January numbers. But one thing is already clear. The market is no longer dazzled by a record; it is watching the rate. And if the rate slows, then even a record becomes a warning in the market's eyes. Samsung may be fine. But the market is now asking: for how much longer?