01. H200 Greenlight to China Meets CPU/GPU Price Hikes
Washington will let NVIDIA ship H200 AI chips to China if the firm hands over 25 % of the revenue, a move framed as balancing national security with corporate profits. Almost simultaneously, AMD and Intel told distributors to raise consumer GPU/CPU list prices 10-20 %, citing AI chips’ swallowing of 3 nm/5 nm wafer starts and CoWoS capacity; memory now accounts for 30-40 % of BOM, and DDR/GDDR prices have doubled in a month. Post-clearance Chinese hyperscalers are bulk-buying HBM3e, tightening GDDR supply further, while Micron’s shuttering of the Crucial brand and tight OEM allocation have pushed spot DRAM up 4× this year. Analysts see CPU/GPU inflation lasting through 1H 2026 and memory adding another 50 % before new capacity loosens the market in 2H.
02. DDR5 Spot Dips 0.25 %, but TrendForce Still Bullish on Q1 Contract Prices
DDR5 spot price edged down 0.25 % on Dec 10 as traders took year-end profit, yet TrendForce says the market remains tight after a 2.5× spike since October and expects contract prices to keep climbing in Q1.
03. AI Ignites “Super-Cycle”—SK hynix Q4 Profit Seen at KRW 16 T
Dec 8, Kiwoom Securities forecasts SK hynix will post a record Q4 operating profit of KRW 16.2 trillion, driven by HBM and commodity DRAM price explosions. Samsung’s DS division is projected above KRW 15 trillion, pushing their combined quarterly profit past KRW 30 trillion—a Korean corporate record. With AI data-center demand still expanding, HBM and NAND prices are rising in tandem; both giants should stay strong in Q1. WSTS now sees the global semiconductor market topping US$1 1 tn by 2026, four years ahead of schedule. Analysts say higher commodity DRAM pricing gives Korean vendors leverage in HBM negotiations, keeping margins fat.
04. Samsung's 4nm yield rebounds, securing a billion-dollar AI chip order.
Dec 9, SemiMedia: Samsung’s 4 nm yield has climbed to 60-70 %, helping it secure an over US$100 M (≈KRW 150 bn) order from U.S. start-up Tsavorite Scalable Intelligence for an “all-in-one OPU” (CPU+GPU+memory) due for 2026 delivery. Though modest, the deal is viewed as a milestone in Samsung Foundry’s recovery; the company also holds Tesla’s AI5 and China crypto-miner 4 nm sockets and will offer discounted 4 nm slots to boost utilization, targeting positive foundry cash flow by 2027.
05. NVIDIA H200 Waiver Benefits Tencent, Alibaba, but Won’t Stop China’s Home-Grow Push
Dec 10, Commercial Times: The Trump administration’s H200 approval is seen as a halfway house between a total ban and shipping Blackwell. Guotai Haitong says the extra compute will help Tencent, Alibaba, and others ramp capex and AI apps, yet leaves China’s long-term substitution goal intact. Guosen notes BAT giants are monetizing AI hardware + models fast, while domestic GPUs are still ramping. A U.S. think-tank warns America’s compute lead could shrink from 10× to 5×, but with domestic cloud demand at ~40 M cards and local share still tiny, ecosystem risk remains.
06. Taiwan’s “top 3” AI Server Makers Double Capacity, U.S. Exports to Surge
Nvidia launched Alpamayo-R1, the world’s first vision-language inference model that gives cars human-like common-sense judgment. Vision specialist Solomon is first to adopt Jetson Thor + Isaac; its robot-inspection kit is already shipping to the Americas, and its AI sales are forecast to jump >70 % this year and next. MediaTek, meanwhile, is co-developing a 3 nm smart-cockpit chip (C-X1) with Nvidia, sampling H2, and is set to add revenue in 2025.