News
AI Memory Shortage: AMD's Lisa Su Identifies High-Bandwidth Memory as AI Chip Supply's Next Cap
10+ hour, 17+ min ago (492+ words) That constraint is now easing. TSMC's Co Wo S output reached approximately 75, 000 wafers per month in 2025, and analysts at Trend Force project further expansion to 120, 000 monthly by the end of 2026 through capacity additions at its AP6, AP7, and AP8 advanced packaging facilities. Nvidia's…...
Embedded substrates draw AI chip interest as packaging turns strategic
2+ hour, 14+ min ago (119+ words) digitimes Embedded substrates draw AI chip interest as packaging turns strategic Keep me signed in - Track financials & stock data of Taiwan tech. - Agibot claims 100% success rate in factory deployment as humanoid race shifts to real-world validation - SK Hynix speeds Yongin…...
AI server boom lifts memory interface chip leader, but DDR5 risks linger
2+ hour, 13+ min ago (146+ words) digitimes AI server boom lifts memory interface chip leader, but DDR5 risks linger Montage Technology, a company that does not manufacture memory chips itself but instead supplies the interface chips linking CPUs and memory, has quietly emerged as one of the…...
Amid A Global Memory Chip Supply Crunch, Is China the Answer?
4+ hour, 42+ min ago (1240+ words) What chips American companies can sell to China has long been a controversial subject. How many chips they can and should buy from China is now rising fast up the agenda too. The reason? A global shortage of memory chips…...
MU Stock May 24: Memory Chip Supercycle Drives AI Boom
9+ hour, 11+ min ago (267+ words) Memory chips are critical for AI infrastructure, storing data for model training and inference. Demand for memory chips far exceeds supply, creating favorable pricing through 2026. New foundries take years to build, ensuring supply constraints persist long-term. Micron's position as a…...
Cheap Optane memory is giving local AI builders a new route.
14+ hour, 6+ min ago (464+ words) Ai / Startup | A Local LLa MA user ran Moonshot AI's Kimi K2. 5 on a workstation using 768 GB of secondhand Intel Optane Persistent Memory and a single RTX 3060. The result is slow but useful enough to show why memory arbitrage may become…...
AI is killing the cheap smartphone. The memory that powers your phone now goes to data centres instead.
12+ hour, 22+ min ago (817+ words) LPDDR prices surged 250% in a year as Samsung, SK Hynix, and Micron redirected wafers to AI. India's sub-$100 smartphone market collapsed 59%....
Kimi K2. 5 runs on RTX 3060 with 768 GB Intel Optane memory at 4 tokens per second
19+ hour, 25+ min ago (272+ words) A Chinese AI enthusiast squeezed a trillion-parameter model onto consumer hardware using second-hand memory DIMMs, and the implications go far beyond the stunt itself. A trillion-parameter AI model just ran on a graphics card that most gamers would consider mid-range....
Kimi K2. 5 runs on RTX 3060 with 768 GB Intel Optane memory at 4 tokens per second " Trading View News
19+ hour, 25+ min ago (240+ words) A trillion-parameter AI model just ran on a graphics card that most gamers would consider mid-range. A Chinese AI enthusiast known as APFrisco demonstrated Moonshot AI's Kimi K2. 5 model, a Mixture-of-Experts (Mo E) large language model with 1 trillion total parameters, running on…...
Samsung Electronics Adds HBM Software Edge, Targets Custom HBM Market
1+ day, 2+ hour ago (127+ words) The Hidden Driver Behind Industry-First HBM4 Mass Production The 'Software Development Kit (SDK)' Emerges as Key Embedding Intelligence into HBM Base Die Spotlighted as Essential Technology for Custom HBM "Samsung Electronics is making full-fledged moves to preempt the market, including expanding…...