AI Bubble

News Feed

Category
AllValuationFundingCapexChips & ComputePolicyModelsMarketsOther
Region
AllUSChinaGlobal

Arm Releases New Compute Subsystem Neoverse CSS N4 for Cloud and Data Centers

At the Arm Everywhere China annual conference, Arm released the Neoverse CSS N4, a new compute subsystem for cloud and data centers. Targeting agentic infrastructure, it offers high configurability, supports 3.8 GHz CPU frequency, LPDDR6 memory, and PCIe Gen 7, based on a 3nm process with up to 128 cores per die. Compared to the previous generation, performance improvements are significant. Hongjun Microelectronics is among the first partners to adopt the technology.

IT Home2026-09-08Chips & ComputeGlobalrel 35Read original ↗

Arm Releases Second-Generation Mobile Compute Subsystem CSS for Mobile 2 with Dual SME2 Engines and Neural Acceleration GPU

At the Arm Everywhere China event in Shanghai, Arm released the second-generation mobile compute subsystem CSS for Mobile 2, a computing platform for agentic AI and AI-native graphics, integrating the C2 CPU cluster, Mali G2-Ultra NX GPU, and system interconnect. The platform features dual SME2 engines in the CPU for significant performance gains and embeds a neural accelerator into the GPU shader cores to support neural graphics workloads.

IT Home2026-09-08Chips & ComputeGlobalrel 35Read original ↗

Samsung Reportedly to Devote Half of Its 4nm Capacity to HBM4 Base Die Production

According to Korean media ZDNet Korea, Samsung Foundry is reserving significant capacity for HBM4 base die production, with about half of its 4nm process (SF4) capacity dedicated to this purpose. As customers demand custom logic integrated into the base die, Samsung's HBM market share reached 38% in Q2 2026, up from 15% a year earlier, while SK Hynix's share fell from 64% to 50%.

IT Home2026-09-07Chips & ComputeGlobalrel 55Read original ↗

Report: 2026 Global and Computing Chip Industry Market Research

EE Times China published a market research report on the 2026 global and computing chip industry, covering market size, technology trends, and competitive landscape. The report notes that demand for computing chips continues to grow with the proliferation of AI applications, but specific figures are not disclosed in the excerpt.

Google News: AI泡沫2026-09-06Chips & ComputeGlobalrel 30Read original ↗

JPR: PC Market Weak in Q2 2026, Discrete GPU Shipments Grow 12.2% Against Trend

According to Jon Peddie Research, discrete GPU shipments grew 12.2% quarter-over-quarter and 14.1% year-over-year in Q2 2026, despite a weak PC market and memory shortages raising costs. Total consumer PC GPU shipments reached 75.5 million units, up 10.4% sequentially and 1.1% annually, driven mainly by laptops. Intel led with 56% market share, followed by Nvidia at 23% and AMD at 21%.

IT Home2026-09-06Chips & ComputeGlobalrel 30Read original ↗

From Power Grid to Computing Power: Toshiba's New Map of Power Semiconductors

This article explores Toshiba's latest moves in power semiconductors, expanding from traditional grid applications to computing infrastructure. It analyzes how Toshiba leverages its technological edge in power semiconductors to supply critical components for AI data centers and other high-performance computing scenarios.

Google News: AI泡沫2026-09-05Chips & ComputeGlobalrel 35Read original ↗

Report: Micron's HBM Monthly Capacity Expected to Reach 100K Wafers by End of 2026, Doubling from Last Year

According to Korean media ETNEWS, Micron is aggressively expanding its HBM memory capacity, with monthly wafer starts expected to rise from 40,000-50,000 at the end of 2025 to 100,000 by the end of 2026, doubling within a year. The expansion focuses on 12Hi HBM4 for AI XPUs like NVIDIA's Rubin GPU, with its share of total HBM output projected to increase from 20-30% at the start of the year to 50% by year-end. To support the ramp-up, Micron is increasing equipment purchases, with equipment arriving at its Taiwan and Singapore fabs.

IT Home2026-09-04Chips & ComputeGlobalrel 70Read original ↗

Samsung and Arm Reportedly Partner to Develop Next-Gen On-Device AI Chips

According to Korean media, Samsung Electronics and Arm have launched a joint development project for on-device AI chips, with Arm providing core AI architecture and design technology, and Samsung handling SoC design and 2nm process mass production. The partnership aims to develop custom chips that can perform computations locally on devices without the cloud, with Samsung's foundry business playing a central role.

IT Home2026-09-04Chips & ComputeGlobalrel 30Read original ↗

TrendForce: Apple iPhone 18 Pro memory procurement costs surge 400%, estimated price increase of 10%-20%

TrendForce reports that Apple's iPhone 18 Pro memory costs have surged nearly 400% year-over-year, putting significant pressure on the overall bill of materials, and new models are expected to see price increases of 10%-20%. Additionally, Apple will adjust its product release cadence, launching high-end models in the fall, including its first foldable iPhone 18 Fold, with a starting price estimated between $2,099 and $2,299.

IT Home2026-09-03Chips & ComputeGlobalrel 30Read original ↗

New Gallium Nitride Transistor Withstands Nearly 4000V, Could Be Used in AI Data Centers and EVs

Researchers at EPFL in Switzerland have developed a new gallium nitride transistor that can withstand nearly 4,000 volts, setting a new record for the technology while maintaining low resistance. The device could be used in high-voltage power electronics for AI data centers, renewable energy systems, and electric vehicles. The research was published in Nature Electronics.

IT Home2026-09-03Chips & ComputeGlobalrel 35Read original ↗

AI Computing Power Drives MLCC Demand, Samsung Electro-Mechanics Secures Largest Long-Term Contract in History

Eastmoney reports that surging AI computing power demand is boosting MLCC (multi-layer ceramic capacitor) demand, with Samsung Electro-Mechanics securing its largest-ever long-term supply contract, suggesting the industry chain's prosperity may continue. This indicates that AI hardware expansion is spreading to more upstream component sectors.

Google News: AI泡沫2026-09-03Chips & ComputeGlobalrel 55Read original ↗

New Frontier in Compute Revolution: Scale-In Optical Interconnect Technology Surges, Bandwidth Overwhelms Traditional Solutions by 10x

The article reports on a breakthrough in Scale-In optical interconnect technology, claiming bandwidth is 10 times that of traditional solutions, potentially driving upgrades in compute infrastructure. The technology is seen as key to meeting AI compute demand, but the article provides no specific investment or market data.

Google News: AI泡沫2026-09-02Chips & ComputeGlobalrel 35Read original ↗

Samsung Galaxy Z Fold 8 Success Triggers Supply Crunch for Key Components

According to Korean media, Samsung's Galaxy Z Fold 8 foldable phone is selling well, causing a supply crunch for its display driver IC (DDI). The chip is designed by Samsung's System LSI and produced by UMC using a 22nm process, but capacity is near its limit and cannot meet Samsung's 'Super Hot Run' request. Samsung even plans to use DDI materials reserved for next year to address the current shortage.

IT Home2026-09-02Chips & ComputeGlobalrel 30Read original ↗