AI that runs on-device — phones, cars, cameras and sensors. Privacy-friendly, latency-free inference outside the cloud.
417 articles
This research identifies an Accuracy-Efficiency Paradox in on-device energy forecasting: highly accurate models can consume more energy than they save. It proposes a Total Cost of Ownership framework that combines inference energy use with battery aging to measure net energy loss.
ArXiv AI · 22d ago
Insta360 introduced Cameraman, a vision for an autonomous photography robot or Camera Agent that can see, understand, act, and deliver finished content. The concept lacks a finalized product form, technical roadmap, or commercialization timeline, while the company's first-half profit fell 94.15% amid sharply higher R&D spending.
虎嗅 · 23d ago
Xiaomi president Lu Weibing unveiled the Xuanjie O100 prototype, an on-device AI terminal designed without a rear camera. The device uses dual-chip computing and active air-cooling to run MiMo at approximately 295 tokens per second.
Pandaily · 25d ago
Xiaomi has unveiled a broader 玄戒 AI chip product family. The company also announced a new wide-folding smartphone, expanding its AI hardware and device portfolio.
Ifanr (爱范儿) · 26d ago
Xiaomi has unveiled the Xuandjie D100, described as China’s first 3nm high-compute AI chip for intelligent driving. The completed chip is scheduled for commercial deployment next year, with capacity for large-scale on-device models.
IT之家 · 26d ago
Anduril is expanding its Seattle-area engineering operation, with regional headcount potentially growing from 560 to 1,000 engineers. The company is developing augmented reality glasses, digital night vision, and edge-computing hardware for the U.S.
GeekWire · 29d ago
NVIDIA demonstrates how developers can build real-time edge AI applications with Holoscan using its CLI, skills, documentation, and AI coding agents. The workflow leverages HoloHub’s reference applications and reusable components to help agents assist with development.
NVIDIA Developer Blog · 31d ago
Ukrainian intelligence claims that Russia’s S-71 “Monochrome” cruise missile contains Nvidia Jetson Orin NX modules. The reported hardware could support AI-enabled terminal guidance, although the claim and the module’s operational role have not been independently verified.
Tom's Hardware · 36d ago
XPeng is changing its vehicle-development and manufacturing approach while positioning AI as a central capability. The company must convince investors that stronger AI can arrive alongside lower hardware costs and visible scale economies.
钛媒体 · 37d ago
BIOSTAR launched the EdgeComp MT-N150, a compact industrial computer built around Intel N150 processors and supporting discrete NPUs from DEEPX and MemryX. It targets edge computing workloads such as intelligent security and smart-city deployments.
IT之家 · 37d ago
ModelBest has filed IPO counseling documents with the Shanghai Securities Regulatory Bureau, appointing CITIC Securities as its advisor. A mid-July funding round valued the company at more than 20 billion yuan, while its MiniCPM models have surpassed 38 million downloads across GitHub and Hugging Face.
Pandaily · 38d ago
面壁智能在半年內完成超過50億融資，估值突破200億，成為端側AI領域的重要獨角獸。文章指出，其高估值仍需要更多商業化成果驗證。
钛媒体 · 38d ago
NVIDIA JetPack 7.2.1 adds Agentic Video Skills and T3000 emulation for Jetson applications. The update targets video-heavy workloads spanning robotics, intelligent video analytics, industrial automation, healthcare, media processing, and remote operations.
NVIDIA Developer Blog · 39d ago
The article revisits Vinci, a 2016 Kickstarter-funded smart hearable that combined cellular connectivity, local storage, voice assistance, and music personalization before the AI hardware boom. It argues that today’s AI earbuds validate Vinci’s vision, but through ecosystem collaboration rather than complete independence from smartphones.
虎嗅 · 39d ago
This roundup highlights major AI developments across consumer hardware, cloud infrastructure, open source, edge models, and regional commerce. Key items include OpenAI’s reported $400 speaker, Apple’s smart-home push, AWS compute shortages, NVIDIA cuFile becoming open source, Google DeepMind’s cyclone model, and advances in vLLM and low-resource language models.
钛媒体 · 42d ago
Nvidia is reportedly seeking a Chinese base-station maker to combine its AI computing platform with wireless network equipment for 6G systems targeting overseas markets. Sources say Nvidia and Shenzhen-based Jiaxian Communications have been in contact since 2025, but the arrangement remains unconfirmed.
TechNode · 44d ago
US company Auterion is providing AI capabilities that allow Ukraine’s low-cost kamikaze drones to track targets autonomously. A $100 million deal will equip 50,000 Ukrainian drones with the US-developed technology.
Ars Technica AI · 47d ago
The new Deepal S05 features the Tianzhu intelligent driving system, powered by dual Horizon J6M chips and end-to-end algorithms. It integrates advanced cabin AI, including a 3nm cockpit chip and large-model voice interaction.
IT之家 · 47d ago
ChangXin Memory Technologies (CXMT) is finalizing validation for its LPDDR6 mobile DRAM, targeting a 12,800 Mbps design rate. The company is positioning itself for mass production by H2 2026, with Apple currently evaluating the technology.
Pandaily · 49d ago
The 'Star Hub' project, a flagship space computing infrastructure in Shanghai, has officially launched its first constellation consisting of three satellites. The project aims to process data directly in space, significantly reducing reliance on ground-based infrastructure.
36氪 · 62d ago
A breakthrough technology enables drones to transmit AI tokens directly via satellite communication. This innovation was recognized with the top award at the World Artificial Intelligence Conference (WAIC).
量子位 · 63d ago
Chinese tech companies are shifting focus from massive, cloud-based models to lightweight AI designed for local execution on smartphones and laptops. This strategy aims to improve processing speed and data privacy by bypassing the need for power-hungry data centers.
SCMP Technology · 65d ago
XPeng is leveraging its successful MONA M03 sedan strategy to develop a new, globally-oriented SUV. The company aims to prove that high-end AI software can be effectively integrated into lower-margin, mass-market vehicles.
钛媒体 · 65d ago
The integration of AI and aerospace technology is accelerating, leading to a surge in the space computing sector. Technological breakthroughs are driving a significant industrial transformation in orbit.
钛媒体 · 65d ago
NVIDIA has introduced two new mass-production modules, Jetson T3000 and T2000, to the Jetson AGX Thor series. These models are designed to optimize power consumption and memory usage for embedded AI applications.
ITmedia AI+ (日本) · 65d ago
The integration of AI and surging DRAM costs are forcing low-end smartphone vendors out of the market. Market leaders like Apple and Samsung are consolidating power while smaller competitors struggle with manufacturing bottlenecks and price pressures.
Computerworld · 68d ago
Google is reportedly set to launch the Tensor G6 chip in the Pixel 11 series, making it the first smartphone processor to utilize TSMC's 2nm manufacturing process. This move positions Google ahead of Apple's upcoming iPhone 18 Pro release schedule.
TechNode · 68d ago
MOVA LINCO raised tens of millions in angel funding to launch its AI-integrated home infrastructure, including an AI router, NAS, and compute box. The products aim to provide a unified 'connection, storage, and compute' architecture for proactive home AI services.
36氪 · 68d ago
Starting July 7, the EU requires all newly manufactured passenger cars and vans to include advanced driver assistance systems. These mandates include driver distraction warning systems and emergency braking capable of detecting pedestrians and cyclists.
The Next Web (TNW) · 72d ago
SpaceX is prototyping a handheld AI device that integrates Starlink connectivity and xAI technology, potentially serving as a distributed computing node for AI inference.
虎嗅 · 73d ago