Below are the most-read DIGITIMES stories from the week of January 16 – February 1, 2026.
Nvidia and Apple bring Intel into the foundry mix
Nvidia and Apple are reportedly planning limited collaboration with Intel for selected 2028-era chips, marking a pragmatic shift toward dual-foundry strategies under US manufacturing mandates. Nvidia's Feynman platform is expected to keep core GPU dies at TSMC while assigning parts of the I/O die and a portion of advanced packaging to Intel, initially via EMIB and later potentially on 14A if yields mature.
Apple is exploring Intel for entry-level M-series chips, reflecting political pressure, cost concerns, and risk diversification rather than dissatisfaction with TSMC. For TSMC, these "non-core" diversions may actually ease regulatory scrutiny and political pressure while strengthening its grip on the most advanced, profitable workloads.
Rising costs push Apple and Qualcomm to rethink TSMC dependence
Apple and Qualcomm are reassessing their heavy reliance on TSMC, not because of technical shortcomings, but due to rising costs, capacity tightness, and geopolitical exposure. Qualcomm has confirmed a renewed dual-track strategy with Samsung, while Apple is evaluating alternatives for lower-tier Mac processors.
With 2nm pricing stretching the limits of consumer electronics economics and HPC demand absorbing much of TSMC's capacity, diversification is emerging as a cost-control and risk-management tool. Still, both companies are proceeding cautiously, limiting alternative foundry use to trials and non-flagship products while keeping their most demanding chips at TSMC.
Lisa Su's hardest decision grounded AMD's comeback
AMD CEO Lisa Su revealed that one of her toughest early calls was scrapping AMD's uncompetitive server roadmap and starting over. This was the defining move that ultimately enabled the company's dramatic resurgence. By abandoning legacy designs, betting early on chiplet architectures with advanced packaging, and committing to TSMC's leading-edge processes, AMD rebuilt its data center business from near irrelevance to roughly 40% server CPU market share.
These decisions not only reshaped AMD's trajectory but also pressured Intel to adopt similar design philosophies, cementing Su's reputation for strategic clarity and execution discipline in the AI era.
Intel hints at a long-term return to DRAM
New research disclosures tied to Intel and US national laboratories have fueled speculation that Intel may be positioning itself for a future return to the DRAM market. Through its Advanced Memory Technology project, Intel has demonstrated next-generation DRAM bonding techniques that promise higher bandwidth, better energy efficiency, and fewer trade-offs between HBM and DDR memory. While commercial timelines remain unclear, the move reflects Intel's search for new growth engines amid AI-driven memory demand and mounting pressure across CPUs, foundry services, and accelerators.
Samsung and peers ignite a sharp NAND flash price upswing
Samsung is reportedly set to more than double NAND flash prices in the first quarter of 2026, joining SK Hynix and SanDisk in aggressive hikes driven by AI-led demand and constrained supply. Server and data center storage needs are tightening the market, with spillover effects seen in DRAM, where RDIMM spot prices continue to surge. Strong financial results from Samsung underscore how the memory supercycle is reshaping industry profitability, with suppliers regaining pricing power as capacity for 2026 is locked in.
Taiwan–US AI memory alliance takes center stage
Micron CEO Sanjay Mehrotra's visit to Taiwan, which was closely followed by Nvidia CEO Jensen Huang, highlighted memory's elevation to a strategic pillar of the AI supply chain. Taiwan's government emphasized deepening cooperation with Micron, including HBM expansion and closer ties with local partners such as PSMC, while Huang publicly reinforced Micron's central role in Nvidia's next-generation AI systems. The coordinated messaging underscored how HBM has become indispensable to AI performance and how Taiwan is positioning itself as a critical hub for AI memory manufacturing.
Jensen Huang reframes Taiwan–US capacity expansion and AI geopolitics
Jensen Huang pushed back against claims that 40% of Taiwan's chip capacity would be moved to the US, stressing instead that global fabs represent net new capacity needed to meet explosive AI demand. He framed overseas expansion as a resilience-driven addition rather than a hollowing-out of Taiwan's role. Huang also dismissed alarmist comparisons about selling AI chips to China, arguing for a balanced approach that protects security while sustaining US technological and economic leadership. Throughout, he emphasized memory as core to Nvidia's AI future.
Article edited by Jack Wu
