KIOXIA just posted one of the most extreme earnings prints in memory history. Q2 revenue hit $15.77B, beating estimates and soaring 453% year-over-year, while operating profit reached $12.68B at an 80% margin. Net profit jumped more than 3,200% as generative AI demand drove sharp increases in SSD average selling prices and bit shipments. The strength is real, but the outlook is more cautious. Q3 guidance came in below consensus, with management pointing to a stronger yen as a headwind even as data-center demand stays robust. The company also announced its first dividend since listing and a policy of returning 100% of surplus cash through progressive dividends and flexible buybacks. Despite the numbers, the stock has struggled. In memory, even explosive AI-driven growth is facing a skeptical market right now.

Samsung Electronics shares fell nearly 5% to around ₩230,000 today despite reporting a record third-quarter operating profit of ₩107.4 trillion (about $80 billion). The KOSPI index dropped roughly 2.7% into the 5,800s, reversing earlier gains above 6,000 and weighed down by the heavyweight stock. Chips drove nearly all of Samsung’s profit, while other divisions lagged. Investors appear disappointed that further shareholder returns beyond the announced ₩30 trillion cash dividend were deferred to a January decision, even as the company outlined large AI-related capital spending. The sell-off highlights how high expectations around returns can outweigh strong earnings.

Samsung Electronics Q3 2026 Earnings Call: Record Profits, AI Boom Visibility to 2027, and Massive Shareholder Returns. [Complete Summery] Memory. On the Q3 2026 call it reported consolidated revenue of 195.25 trillion won and operating profit of 107.44 trillion won. Fourth straight all-time records. Revenue up 126.9 percent year-on-year. Operating profit up 783.1 percent. Operating margin roughly 55 percent. Device Solutions did the work. It delivered 147.1 trillion in revenue and 106.5 trillion in operating profit. That is about 99 percent of the entire company’s profit. Memory revenue alone jumped 427% to 140.7 trillion, driven by HBM and server DRAM as hyperscalers kept building. It sees the tightness lasting. Prolonged AI investment plus supply constraints should keep conditions tight into 2027. It expects 2027 results to beat 2026. Demand is forecast to far outpace supply. The response is more HBM4, mass production of HBM4E, and a shift toward high-performance DDR5. HBM4 revenue more than tripled quarter-on-quarter. It now makes up about 60 percent of total HBM sales. It has hit its goal of HBM share matching its overall DRAM share. For 2027 it has already locked full planned volumes with customers at target prices. Sold out. Customers still want more. Foundry is moving too. It expects 2-nanometer order wins this year to more than double. New projects from large cloud and AI/HPC customers in Q2 and Q3. AI and HPC products are projected to be more than half of foundry revenue next year. Growth includes the Taylor plant ramp in Texas. Then the contrast. Device eXperience (phones and appliances) saw sequential revenue rise to 52.1 trillion. It still posted an operating loss of about 800 billion. Second straight quarter of losses that size. Higher memory and component costs kept squeezing the margin. On capital expenditure, Samsung plans to spend 84.4 trillion won on production facilities this year. Of that total, 76.8 trillion won is allocated to the Device Solutions semiconductor division and 3.9 trillion won to Samsung Display. The board approved a Q3 cash dividend of 4,601 won per share, about 30.03 trillion total. That brings executed or confirmed returns this year to roughly 35 trillion of the 90-to-110-trillion program. The rest goes to more dividends plus buybacks and cancellations. It will actively consider preferred shares in the buyback mix. Next multi-year policy expected at the late-January board meeting. The same memory boom that is generating almost all the cash is also the reason finished-product margins stay under pressure. Visibility runs through 2027. The phone line is still paying for it.

Samsung Electronics announced that it has completed preparations for mass production of 10th-generation V-NAND products with bonding technology in the third quarter, and plans to commercialize a 256-terabyte (TB) SSD with 32-layer stacking technology applied to 9th-generation V-NAND-based QLC products early next year.

Korea’s financial watchdog says Solidigm’s Nasdaq listing looks like a dual listing. It plans to raise the concern with SK Hynix. Chosun Biz reports that Financial Supervisory Service Governor Lee Chan-jin told lawmakers on the 19th that Solidigm’s planned US listing raises a dual-listing issue. He said both the Financial Services Commission and his agency share that view and are monitoring the matter. He is waiting for an official interpretation on whether a securities registration statement is required to protect existing Korean shareholders. If there is a way to convey the concern to SK Hynix, he said he will do so. A lawmaker called the Nasdaq plan a typical case of Korea’s long-running dual-listing problem expanding overseas. He noted that SK Hynix explained the potential negative effects in detail to the US Securities and Exchange Commission but gave no such explanation to Korean shareholders. $SKHY If the watchdog already sees a dual-listing problem and is waiting on an official reading, does the next step force more disclosure for Korean holders, or does the Nasdaq path still move ahead unchanged?

Samsung is cutting mid-to-low-end phone output again. Memory prices are not the only reason. As per Korean media reporting, Samsung has lowered its fourth-quarter mid-to-low-end smartphone production plan. Rising memory prices are the main driver, but printed-circuit-board material is also tight. Japan’s Nitto Boseki has shifted capacity toward higher-margin T-glass used in AI semiconductor substrates, reducing E-glass used in smartphone boards. That has raised prices and made supply harder. Samsung has cut related component orders as a result. The lower-margin Galaxy A series is the main target. Full-year smartphone shipments are now seen around 230 million units, below the prior 240 million expectation. Updated fourth-quarter monthly targets total about 24.2 million units, down 23% from the earlier 31.4 million plan. Flagship plans for the S27 series early next year stay higher, and foldable volume is still set to rise. The Mobile Experience division posted a 735 billion won loss in the second quarter. Some broker estimates put the full-year loss around 400 billion won. Mid-to-low-end base models have already seen price increases of more than 100,000 won versus the prior generation. AI demand is lifting memory profits while squeezing the phone business that uses the same memory. #Samsung# If memory and board material both stay tight, does Samsung protect overall profit by shipping fewer mid-range phones, or does that open share for rivals that can still absorb the cost?

WHY so much hype about CXMT? CXMT is guiding third-quarter revenue up 589% to 619%. The stock reaction treated that as a technology event. The yield question is still open. It forecasts 114.7 billion to 119.7 billion yuan of revenue and a swing to 53.3 billion to 54.8 billion yuan of net profit. Chairman Zhu Yiming proposed a 1 billion to 2 billion yuan share buyback. Elon Musk has said China will produce an enormous amount of memory by 2028. Apple is reported to be testing its memory. The 4F² structure can cut cell area 33% versus the current 6F². Rosenblatt and Mizuho say the process level and yield are not yet verified, so the part cannot be compared directly with Samsung’s 2028 mass-production target. Samsung fell 17.8% and SK Hynix 24.3% last week. Separately, Samsung has accepted NVIDIA’s request to use TSMC’s 3nm process for the base die of custom NVHBM. T he core die stays on sixth-generation DRAM (1c). That is the first time it has partnered with TSMC from the HBM manufacturing stage. It keeps its own foundry for standard HBM and for customers such as AMD. NVHBM is based on the HBM4E structure. A 600% revenue print is a volume and pricing print. It is not a qualification of the 4F² cell. $MU $SKHY $DRAM #cxmt#