Below are the most-read DIGITIMES Asia stories from the week of February 2-9, 2026.
China's memory industry moves upmarket
China's leading memory makers, CXMT and YMTC, are decisively abandoning their earlier low-price strategies as they pivot toward higher-end DRAM and NAND, with pricing now approaching or even exceeding South Korean competitors. Contrary to market rumors of aggressive DDR4 dumping, CXMT has largely shifted resources to DDR5 and LPDDR5, maintaining only minimal DDR4 output for long-term customers. Improving yields, rising DRAM prices, and a sector upcycle in the second half of 2025 have sharply improved profitability, putting CXMT on track for full-year net profits well ahead of earlier forecasts. With major capacity expansions not expected to meaningfully contribute until 2027, memory supply tightness is seen as structural through 2026–2027, reinforcing supplier pricing power as AI-driven demand continues to climb.
The rise of the agentic enterprise
Salesforce characterizes 2026 as a turning point for "agentic enterprises," where autonomous AI systems move beyond task automation toward shared accountability for business outcomes. The company highlights three defining trends: the rise of multi-agent systems that function as an orchestrated digital workforce; AI-driven customer experiences becoming a core element of brand identity; and AI evolving into a "digital immune system" capable of continuous cybersecurity monitoring and self-healing responses. While CIO adoption of AI has surged, Salesforce notes that data governance and trust remain key bottlenecks. The vision is particularly relevant for Taiwan, where cyberattacks are among the most frequent in Asia-Pacific, underscoring the appeal of AI systems that can respond faster than human teams.
Germany turns to Korea
Germany is intensifying semiconductor cooperation with South Korea after Intel halted plans for major European fabs, creating gaps in the region's advanced chip ambitions. Delegations from Saxony and Saxony-Anhalt are visiting South Korea to court chipmakers and suppliers, leveraging existing ties that already link Korean firms to Germany's R&D ecosystem in Dresden. The move complements Germany's support for TSMC's ESMC project and coincides with EU approval of substantial state aid for new fabs by GlobalFoundries and X-FAB. Together, these efforts reflect Europe's push to diversify supply chains and reduce strategic dependence as global semiconductor competition tightens.
SK Hynix advances towards Nvidia's HBM4 qualification
SK Hynix has made significant progress in Nvidia's quality testing for sixth-generation HBM4, strengthening expectations that it will be a stable supplier for Nvidia's Rubin GPUs. After addressing earlier circuit issues, revised samples delivered in January 2026 have shown bandwidth nearing Nvidia's 11Gbps target while maintaining performance under stringent stress tests. Although purchase orders are not yet finalized, the results point toward a potential transition from sampling to mass production as early as late first quarter or early second quarter of 2026, even as Samsung prepares to ship its own HBM4 products in February.
HBM4: yield over speed
Samsung Electronics and SK Hynix both enter 2026 with production-ready HBM4 technology, but industry consensus holds that yields will determine market leadership more than headline performance. Samsung has demonstrated higher peak speeds using 1c DRAM and advanced packaging, while SK Hynix carries a yield advantage built on its mature HBM3E experience. Analysts expect SK Hynix to retain a majority share of initial Nvidia HBM4 orders, reflecting stronger cost control and yield stability, while Samsung seeks to use HBM4 to rebalance profitability amid weaker results in its foundry and system LSI businesses. Policy risks, including potential US tariffs, add another layer of uncertainty to the race.
Risk production becomes new HBM norm
As Nvidia's HBM4 qualification nears completion, SK Hynix is accelerating 1b DRAM capacity expansion to ensure supply for next-generation AI accelerators launching in 2027. Both Samsung and SK Hynix have begun "risk production," committing wafers before final certification to overcome long lead times and tight capacity. While this approach carries inventory and defect risks, it reflects strong confidence in commercialization and the strategic importance of meeting Nvidia's schedules. Multiple speed variants are being tested, suggesting that supply will balance top-tier performance with manufacturability constraints.
Samsung and SK Hynix upgrade to "super suppliers"
Under an intensifying global memory crunch, Nvidia is reportedly pressing Samsung for earlier HBM4 deliveries, even before final validation is complete. With conventional memory supplies tightening as capacity shifts toward HBM, bargaining power is swinging sharply toward suppliers. Executives from Nvidia, Tesla, and Intel have all warned that memory shortages could persist for years, making access to HBM a decisive factor in the AI race. As a result, Samsung and SK Hynix are emerging as pivotal gatekeepers of AI progress, controlling a resource that increasingly rivals GPUs themselves in strategic importance.
Article edited by Jerry Chen
