Memory makers just can't churn out their DRAM fast enough. On the heels of an AI-driven shortage, SK Hynix on Tuesday announced a new 19 trillion Korean won (about $13 billion) advanced packaging and test facility in South Korea that could offer some relief - just not for consumer products like laptops and phones. The South Korean memory giant unveiled the new site, dubbed P&T7, which will be located at the Cheongju Technopolis Industrial Park in Chungbuk, South Korea.
DRAM contract prices have increased 171 percent year over year, according to industry data. Gerry Chen, general manager of memory manufacturer TeamGroup, warned that the situation will worsen in the first half of 2026 once distributors exhaust their remaining inventory. He expects supply constraints to persist through late 2027 or beyond.
The ongoing AI boom is signaling good fortune for Samsung, which currently holds the title of leading chipmaker. Samsung has released its Q3 earnings guidance and is anticipating its largest quarterly profit since 2022. Revenues are expected to rise to KRW 86 trillion ($60 billion) while profits are now estimated at KRW 12.1 trillion ($8.5 billion). These figures represent 32% and 8.7% bumps compared to last year.
PC memory prices are set to rise as the major suppliers allocate manufacturing capacity to the more lucrative server DRAM and HBM instead amid reports of tightening supplies. Memory prices are set for an increase in Q4 of 2025, according to market watcher TrendForce, which points the finger at the three top DRAM makers - Samsung, SK Hynix, and Micron Technology.
Nvidia on Tuesday unveiled the Rubin CPX, a GPU designed specifically to accelerate extremely long-context AI workflows like those seen in code assistants such as Microsoft's GitHub Copilot, while simultaneously cutting back on pricey and power-hungry high-bandwidth memory (HBM). The first indication that Nvidia might be moving in this direction came when CEO Jensen Huang unveiled Dynamo during his GTC keynote in spring. The framework brought mainstream attention to the idea of disaggregated inference.
