KB Securities said on the 29th that SK hynix is expected to see improved results in 2027 as the memory supply shortage deepens and high bandwidth memory (HBM) prices more than double. In particular, it projected operating profit in 2027 would reach 454 trillion won.
It maintained a "Buy" recommendation and raised the target price to 3.8 million won from 3 million won. The previous trading day's closing price of SK hynix was 2,289,000 won.
Kim Dong-Won, head of research at KB Securities, said, "We judge that the memory supply-demand imbalance in 2027 will see a further deepening of the shortage compared with 2026, leaving ample room for additional price increases," adding, "HBM prices are expected to rise more than 100% year over year."
Considering clients' memory demand for 2027, it expected next year's demand growth rate to reach 20%. It also saw a high likelihood that this year's shortfall will be deferred into 2027 and translate into additional demand.
For HBM, it analyzed that price negotiations are proceeding smoothly, reflecting a narrowing margin gap with commodity DRAM, whose operating margin has surpassed 80%.
KB Securities raised its operating profit estimates for SK hynix for this year and next to 280 trillion won and 454 trillion won, respectively. The reason, it said, is that as of the second quarter, clients' memory demand fulfillment rate is only 50%, and DRAM and NAND prices are expected to exceed market expectations. It projected that the memory semiconductor market will remain in a supply shortage phase for at least two years through 2028.
Kim said, "The newly blossoming agentic artificial intelligence (AI) is expected to increase token usage sevenfold over the next year, which will drive a long-term surge in AI server demand and larger memory content per system," adding, "Because new capacity additions are concentrated on expanding HBM production, new supply of commodity memory is effectively possible only through process conversions."
Kim also explained that the memory market is being re-evaluated as a structurally growing industry driven by AI investment and the expansion of AI applications. This structural change in memory demand, it said, is leading to upward revisions in earnings and higher valuations, and ultimately will serve as the key driver of share price gains.
He said, "Global Big Tech is betting everything on securing a stable supply of memory semiconductors, a scarce strategic asset, to execute annual AI infrastructure investments of 1,000 trillion won," adding, "On Nvidia's Vera Rubin platform, slated for release in the second half of this year, memory cost in value terms is expected to expand to more than five times that of Blackwell."
It further projected that memory's share of total cost on Nvidia's graphics processing unit (GPU) platform will triple from 8% on the previous generation Blackwell to 25% on Vera Rubin.