AI-driven chip shortages, rising memory costs and intensifying foundry competition dominated the technology sector, with Intel reviving DDR4 platforms, Qualcomm raising prices, Samsung challenging TSMC at 2nm and Beijing accelerating domestic chip adoption. Below are the most-read DIGITIMES stories from the week of July 27- August 2, 2026.
Intel's reported DDR4 CPU revival expected to boost PC demand in 2H26
Intel is reviving DDR4-compatible CPUs, with limited shipments starting in June and volume deliveries expected in September, to offset high DDR5 prices and support China's Singles' Day and Double 12 PC demand. The plan covers 10th Gen Comet Lake, 12th Gen Alder Lake, and 13th- and 14th Gen Core chips.
Apacer CEO C.K. Chang expects DDR4's lifecycle to extend and supply to tighten in the second half, with prices already above US$4 per Gb and Taiwanese suppliers including Nanya Technology regaining pricing power.
Unimicron says Intel's EMIB-T will not reach mass production until 2027
TSMC's tight CoWoS capacity is strengthening demand for Intel's lower-cost EMIB-T platform, though Unimicron expects true mass production only in 2027. EMIB-T can cut packaging costs by 40% to 50% by avoiding an expensive interposer, attracting ASIC, CPU and AI-chip customers that may remain for three product generations.
Unimicron, Ibiden, and Shinko Electric Industries are each targeting an initial 50% yield. EMIB-T represents under 10% of Unimicron's investment, with dedicated equipment accounting for only 15% to 20% of that amount.
Qualcomm tells customers chip prices will rise by double digits from September
Qualcomm plans a double-digit chip price increase from September 1 after exhausting options to absorb higher supplier costs, Bloomberg reported. The rise will affect products shipped after the deadline and could lift costs for Samsung, Xiaomi and other Android brands using Snapdragon chips across smartphones, tablets, wearables, vehicles and networking devices.
The move reflects wider pressure from memory shortages, AI-driven semiconductor demand and tight TSMC capacity, leaving device makers to absorb costs, raise prices or adopt lower-tier processors.
Apple, Micron's clash over CXMT puts White House in the middle
Apple and Micron are lobbying the Trump administration over Apple's proposed use of CXMT and YMTC memory in non-US devices. Apple says Chinese chips could lower costs, while Micron warns they could weaken US producers. The clash comes as Micron posts an 81% second-quarter 2026 margin and plans more than US$250 billion in US investment by 2035.
Samsung foundry to double 2nm design wins in 2026 on Broadcom talks
Samsung expects 2nm project wins to more than double in 2026, with Broadcom and major cloud and AI/HPC customers in talks or design work. Advanced nodes of 8nm and below are fully utilised, AI/HPC could exceed 30% of foundry applications, and advanced nodes over 50% of revenue.
Taylor Fab 1 starts in 2026, Taylor Fab 2 construction begins this year for 2030 production, while rising 1.4nm inquiries and better yields, pricing and utilisation could support a near-term profit turnaround.
China chip substitution accelerates as chill effect looms
Beijing is pressuring Chinese AI companies to adopt domestic chips, with cloud AI facing the strongest mandate despite performance trade-offs. Elsewhere, Qualcomm, MediaTek, Nvidia and NXP remain entrenched in Chinese EVs, while Taiwanese suppliers retain technical and supply advantages. Sources say geopolitical pressure remains the main substitution force, with broader displacement still manageable.
Samsung's HBM4 base die is the foundry weapon TSMC and Intel can't copy
Samsung's foundry recovery increasingly rests on AI-driven capacity overflow from TSMC, with 2nm design wins set to more than double and Broadcom, cloud and HPC customers evaluating its leading-edge nodes. Its strongest advantage is HBM4 integration across memory, logic and packaging, while Taylor, Texas targets the same diversification demand as Intel 18A. Yet Samsung remains near one-tenth of TSMC's market share, making second-half 2nm yield execution the decisive test.
Article edited by Jerry Chen