Samsung Shares Fall 6% Despite 15-Fold Jump in Chip Profit as SK Hynix Raises the Bar

Samsung Electronics reported a 15-fold increase in quarterly operating profit from its semiconductor business, driven by surging demand for chips used in artificial intelligence infrastructure.

However, its shares fell sharply on Thursday as investors looked for stronger guidance on the duration of the memory-chip shortage and clearer commitments to shareholder returns.

The market reaction highlighted the growing gap between Samsung’s earnings recovery and investor expectations. While the South Korean technology giant delivered substantial profit growth, rival SK Hynix had set a higher benchmark a day earlier by forecasting a prolonged supply shortage and announcing plans to return approximately $75 billion to shareholders this year.

Samsung shares initially rose before reversing course, falling as much as 6.2% in morning trading. SK Hynix shares gained 0.7%, suggesting investors were rewarding not only earnings performance but also confidence in future demand and the distribution of cash generated by the AI boom.

“SK Hynix did quite a lot more than expected yesterday … it went beyond market expectations, but Samsung Electronics hasn’t really changed anything from the guidance it gave before,” said Lee Min-hee, an analyst at BNK Investment & Securities.

Samsung reported operating profit of 107.4 trillion won ($80 billion) for the July-to-September period, compared with 12.2 trillion won a year earlier. The result was in line with its estimate. Its semiconductor division generated 106.5 trillion won in operating profit, accounting for almost all of the group’s total, although the figure fell slightly short of the 107.3 trillion won expected by analysts surveyed by LSEG.

The results demonstrate how strongly AI-related demand is reshaping the memory-chip industry. They also expose Samsung’s challenge in translating that demand into a stronger market valuation when competitors offer more aggressive forecasts and clearer capital-return commitments.

SK Hynix said its dynamic random-access memory (DRAM) and high-bandwidth memory (HBM) chips were sold out through 2029, forecasting that supply constraints could persist until 2030. The company’s assessment helped ease concerns that the rapid expansion of AI infrastructure might eventually weaken and undermine the profitability of memory-chip producers.

HBM chips are crucial for AI accelerators because they provide the high-speed memory access required to process large volumes of data. As cloud providers and other technology companies expand their computing capacity, demand for these chips has become a critical source of growth for memory manufacturers.

Samsung also reported strong demand, saying its HBM supply for next year had been sold out and that major customers were asking for additional volumes. However, its forecast for the wider market was more conservative than SK Hynix’s.

Samsung expects the memory-chip market to remain tight through 2028, arguing that the construction of new fabrication plants will take time despite a sharp increase in investment. The company said AI infrastructure capital expenditure would reach historic levels in 2027 as the ecosystem expands, with AI-driven demand expected to grow substantially faster than industry supply compared with 2026.

The difference in outlook matters because investors are trying to determine how long the current pricing environment can support elevated earnings. SK Hynix’s projection of shortages extending into 2030 provides a longer period of potential pricing power than Samsung’s assessment, even though both companies expect demand to remain strong.

“Samsung’s decline was likely attributable to its outlook on how long the memory shortage would last, which came across as marginally less bullish than SK Hynix’s commentary yesterday,” said Jing Jie Yu, an analyst at Morningstar.

Samsung’s investment plans indicate that it expects the AI-driven expansion to continue. Capital expenditure for its chip division rose 62% this year to 76.8 trillion won, with further increases expected next year. Yet additional capacity requires substantial upfront spending and cannot immediately resolve supply constraints, creating a delicate balance between capturing current demand and avoiding excessive investment if the market eventually cools.

The broader investment environment is also becoming more complicated. Concerns about how technology companies will finance expensive AI infrastructure, alongside intensifying competition from China, have weighed on chip stocks. Alphabet, Google’s parent company, reported a sharp acceleration in cash spending on Wednesday as it invested heavily in AI infrastructure, reinforcing investor scrutiny of the costs associated with the industry’s expansion.

Shareholder Returns and Other Businesses Weigh on Sentiment

Samsung’s decision to postpone details of its next shareholder-return policy was another source of disappointment. The company had previously indicated that shareholder returns this year would total between 90 trillion and 110 trillion won, including 30 trillion won in third-quarter dividends.

On Thursday, it said it would finalize its next policy in January, leaving investors without an immediate announcement of additional share buybacks or dividend payments.

“The market appeared to expect an additional buyback and/or dividend announcement, but the company deferred the decision until late January, potentially disappointing near-term expectations,” said Sanjeev Rana, a senior analyst at CLSA.

It comes when investors are increasingly assessing AI-related earnings alongside the cash companies can return to shareholders. SK Hynix’s planned $75 billion distribution gave investors a concrete measure of how it intends to share the benefits of elevated demand. Samsung’s decision to delay its next announcement left that comparison unresolved.

Samsung is also attempting to reduce its exposure to volatile spot-market pricing by securing longer-term supply agreements. It said it had signed such contracts with 10 major customers and was negotiating with another 10, aiming for these agreements to account for 60% to 70% of total production capacity by 2030.

The company is extending these arrangements beyond hyperscalers and other AI-related customers to include mobile-device and automotive manufacturers. Longer contracts could improve revenue visibility and help Samsung plan production, although their financial benefits are tied to pricing terms and the extent to which they protect margins when market conditions change.

Its foundry business remains another challenge. Analysts said losses in the division continued to weigh on semiconductor earnings, even as Samsung reported improved revenue from logic dies used in HBM chips and new orders for advanced 2-nanometer chips from hyperscalers and other AI customers.

Samsung plans to begin mass production next year at its Taylor, Texas, facility, which will manufacture chips for Tesla. It is also considering building another fabrication plant to meet demand for advanced 2-nanometer processors. These initiatives could strengthen its position in contract manufacturing, but the business requires heavy investment and must compete against established foundry leaders.

Meanwhile, higher memory-chip costs squeezed margins in Samsung’s mobile business, which recorded an operating loss of 800 billion won. The result illustrates an important complication in the current boom, where higher memory prices benefit chip suppliers but increase costs for businesses that purchase those components for smartphones and personal computers.

Samsung’s latest results confirm that AI demand is delivering substantial earnings growth. But Thursday’s share-price decline suggests that investors are looking beyond the headline profit increase.