Memory to Make Up 60% of Chip Sales as AI Boom Lifts Samsung, SK Hynix Through 2030

Research Firms Say Memory Boom Has Room to Run

Demand Is Spilling Beyond HBM Into DRAM and NAND

Samsung, SK Hynix Poised to Benefit From Multiple Tailwinds

"Memory Supercycle Could Last Until 2030"

Heavy investment in artificial intelligence infrastructure is lifting not only high-bandwidth memory, or HBM, but also conventional DRAM and NAND flash, pushing the global memory market into an unusually strong upcycle. Major market-research firms project memory will lead this year's surge in semiconductor growth. Demand for high-value AI memory is climbing rapidly, while shortages and price increases are spreading to mainstream products. That is strengthening expectations that Samsung Electronics Co. and SK Hynix Inc. will post strong earnings for an extended period.

Memory Seen Accounting for 60% of Chip Market as DRAM and NAND Surge

A roundup of forecasts from SigmaIntel, Counterpoint Research and Omdia on Aug. 10 showed the global semiconductor market is set for steep growth this year, led by memory. SigmaIntel forecasts global semiconductor revenue of about $1.587 trillion this year, up 107% from a year earlier. Memory revenue alone is expected to reach about $960 billion, up 290%, accounting for roughly 60% of total chip sales.

DRAM and NAND flash are both driving the expansion. SigmaIntel forecasts DRAM revenue will rise more than 300% this year from a year earlier, while NAND flash revenue will increase by nearly 250%. The HBM market alone is projected to exceed $100 billion.

Other research firms have offered similar projections. Omdia raised its forecast for semiconductor revenue growth this year to 94.1% from a year earlier, saying AI demand is outstripping global semiconductor supply capacity. It also expects memory chips to account for more than half of total semiconductor revenue.

Counterpoint Research forecasts the global memory market will quadruple this year to 1,500 trillion won ($1.08 trillion) from a year earlier, and expand to 2,100 trillion won ($1.51 trillion) next year. The firms use different market definitions and methodologies, but their conclusions point in the same direction: AI infrastructure spending is overwhelmingly driving memory-market growth.

Demand is also shifting quickly toward servers. Counterpoint forecasts server products will account for 56% of total memory revenue this year, up from 37% last year. That implies AI data-center spending is reshaping the market so servers generate more than half of all memory sales.

HBM Expansion Tightens Commodity DRAM Supply as AI Demand Reaches NAND

The main pillar of the memory boom is supply constraint. As Samsung, SK Hynix and Micron Technology Inc. aggressively expand HBM capacity, wafer supply for DDR products has fallen by a double-digit percentage. With DDR demand for AI servers also rising, SigmaIntel said supply across the DDR lineup has become exceptionally tight.

HBM is more complex to manufacture than standard DRAM, and large-scale production capacity is effectively limited to Samsung, SK Hynix and Micron. HBM is being deployed in large volumes in AI accelerators from Nvidia Corp., Advanced Micro Devices Inc., Intel Corp. and Google, and supply growth is failing to keep pace with demand.

The push to expand HBM output is, paradoxically, worsening shortages in commodity DRAM and driving prices higher. SigmaIntel projects average selling prices for DDR in the third quarter will reach four to five times the level of the third quarter of last year.

Unusual price inversions are already appearing in some products. Counterpoint said surging server demand, a production mix focused on higher-value products and supply constraints have pushed the per-gigabit price of some commodity DRAM above that of HBM. In other words, some mainstream products now carry higher unit prices than HBM, despite HBM's more complex manufacturing process and higher production cost.

The AI-driven memory windfall is also spreading to NAND flash. Demand for enterprise solid-state drives, or eSSDs, used to store data in large AI servers is rising so quickly that demand for eSSDs for AI servers could surpass NAND demand from consumer electronics.

Supply bottlenecks also appear difficult to resolve quickly. Omdia said shortages are emerging not only in HBM, but also in advanced packaging and leading-edge process technologies. TSMC's advanced packaging lines in Taiwan are running at high utilization, while long lead times for related equipment are making rapid capacity expansion more difficult.

Omdia expects bottlenecks across HBM, advanced packaging and leading-edge manufacturing to persist at least through 2027. Counterpoint also forecasts the memory upcycle will continue through the first half of 2027. It left open the possibility of price corrections after the second half of 2027, when new facilities begin ramping more fully and supply starts to increase.

Samsung, SK Hynix Gain Multiple Tailwinds as Cycle Looks Set to Last Longer

The market environment remains favorable for Samsung and SK Hynix, which produce HBM, commodity DRAM and NAND flash. They stand to benefit both from HBM's high profitability and from rising prices for commodity DRAM and NAND caused by supply shortages.

Samsung ranked first in the global DRAM market by second-quarter revenue, with a 39% share. Rising commodity DRAM prices and a larger HBM share helped lift earnings. SK Hynix's DRAM revenue rose 214% in the same period from a year earlier.

The main variable is competition from Chinese memory makers. Counterpoint said second-quarter DRAM revenue at China's CXMT surged 716% from a year earlier. Revenue at Taiwan's Nanya Technology Corp. climbed 690% over the same period. CXMT in particular is rapidly expanding in commodity DRAM while Samsung, SK Hynix and Micron focus capacity on HBM and other higher-value products. Even so, most forecasts indicate higher Chinese output will not be enough to ease the global memory shortage anytime soon.

Goldman Sachs also said this memory cycle could prove stronger and longer-lasting than past ones. The bank cited AI computing demand growing faster than memory supply, along with structural changes such as an increase in long-term supply agreements. It expects the memory shortage to persist through 2030.

Timothy Moe, Goldman Sachs's Asia-Pacific strategist, said the market is concerned about the memory cycle, hyperscaler capital expenditure, capital-market financing and intensifying competition. Still, those factors are not enough to undermine the long-term boom scenario.

Hong Min-seong, Hankyung.com reporter mshong@hankyung.com