01. DRAM Prices Skyrocket 500% as Global Shortage Spreads
Nov 27 – Semiconductor industry outlet SemiMedia reports global DRAM prices have jumped 500% in recent weeks, with NAND flash up 100%. PC builder CyberPowerPC will raise memory-module prices in December and warns the rally could last into 2026. U.S., Japanese, and Taiwanese retailers are limiting sales or switching to spot pricing; 64 GB DDR5 kits now cost almost as much as a PS5 Pro. Analysts blame surging AI-infrastructure demand: Samsung, SK hynix, and Micron are prioritising high-margin AI chips, leaving legacy DRAM capacity critically tight. Dell, HP, Lenovo, and Asus have all flagged soaring costs and are stockpiling inventory. Counterpoint expects memory prices to rise another 50% in Q2 2026 and warns that shortages could persist for a decade.
02. China, EU Reopen Nexperia Talks to Prevent New Chip-Supply Shock
Nov 27 – China’s Commerce Minister Wang Wentao and EU Executive Vice-President Šefčovič held a video call to defuse the Nexperia dispute. Beijing urged the Netherlands to present a long-term fix; Brussels pledged to keep easing tensions. Both sides agreed to let Nexperia’s Dutch arm and Chinese entities lead negotiations to restore supply-chain stability and will keep talking on export controls.
03. AMD to Raise GPU Prices at Least 10% Again in 2026, Citing Memory Surge
Nov 26 – SemiMedia reports AMD has informed partners it will hike GPU prices at least 10% next year, the second increase of 2025, after memory costs spiked almost 200%. The flagship RX 9070 XT’s Black-Friday dip to USD 599 may be short-lived. Nvidia has also delayed the RTX 50-Super series due to component shortages, prioritising high-margin data-centre cards and keeping consumer-GPU prices elevated.
04. Panasonic to Lift Tantalum-Capacitor Quotes; Distributors, agents, Nippon Chemi-Con, and Podak to Benefit
Nov 27 – Panasonic confirmed across-the-board increases on 30–40 tantalum-capacitor lines. Distributors Nichiden-Bodo (solid caps 24% of revenue, also Kemet/Nippon Chemi-con) and Pota (full industrial, AI-server and power-management portfolio) are poised to pocket the upside. Investors hope the pair can replicate last year’s “Miwangshi model” and turn price hikes into profit.
05. Musk: “We’ll Mass-Produce a New Chip Every 12 Months.”
Nov 25 – Tesla chief Elon Musk wrote on X that his AI-chip team will expand to ship a new silicon design every 12 months, outpacing all rivals combined. AI5 is nearing tape-out, AI6 has already started, with volume production targeted for mid-2027. Musk attends weekly reviews and is recruiting global AI/IC talent to shrink iteration cycles and production lead times.
06. Sovereign-Cloud Boom to Push AI-Server Shipments Above 20% Growth in 2026
Nov 27 – TrendForce says North American hyperscalers will ramp capex next year while countries build “sovereign clouds”, lifting AI-server shipments by more than 20%.
·Rising rivalry: AMD rolls out rack-scale packages to win cloud deals; U.S. hyperscalers accelerate in-house ASICs; ByteDance, Alibaba, Baidu, and Huawei fast-track custom AI chips.
·Cooling upgrade: With single-chip TDP soaring, liquid-cooling penetration will hit 47% in 2026, led by cold plates.
·Transfer bottlenecks: HBM, optical interconnects, and CPO are now top R&D priorities for GPU vendors and cloud builders crafting next-gen smart-compute infrastructure.