84 articles
Mobilint, a Korean AI chip startup, is scaling production of neural processing units designed for inference tasks rather than training, targeting surveillance,...
Aviation is embracing on-device AI to manage complex air traffic without relying on cloud infrastructure, marking a shift toward local intelligence that...
Google is launching three tiers of AI-powered smart glasses focused on productivity and real-world utility, directly challenging Meta's social-first approach...
OpenAI is developing an AI-first smartphone with Qualcomm and MediaTek, targeting 300-400 million annual shipments by 2028.
Skymizer's new HTX301 chip runs massive 700-billion-parameter AI models on a single card at just 240 watts, eliminating the need for expensive GPU clusters and...
Researchers unlock Apple's Neural Processing Units for advanced AI models, cutting response times by up to 5.5x and slashing power consumption by 7.37x on...
Edge AI developers face a critical bottleneck: running large language models on devices with limited memory.
Neural Processing Units are replacing GPUs for edge AI inference, delivering 10-20x faster performance at a fraction of the power.
First responders are using AI-powered glasses and wearables to make faster, safer decisions in life-or-death situations.
Memory shortages are forcing a structural shift in chip manufacturing, with 82% of producers raising prices and only 14% expecting relief soon, threatening the...
Testing four compact AI models on a Raspberry Pi reveals a stark performance gap. TinyLlama responds in minutes while DeepSeek-R1 takes nearly two hours,...
AI models are outpacing the chips designed to run them. Researchers are building adaptable hardware and ultra-low-power processors that let AI work locally on...
Occupancy sensors running AI locally eliminate privacy risks and latency delays. Here's why enterprises are abandoning cloud-dependent systems for edge...
Rokid launches AI Glasses Style with support for multiple AI models including Google Gemini, ChatGPT, and DeepSeek, positioning open ecosystems as an...
Apple is testing four frame designs for its N50 smart glasses launching in 2027, challenging Meta's 7 million Ray-Ban units sold in 2025.
Meta's Ray-Ban smart glasses hit their lowest price ever at $224.25, signaling a shift toward mainstream AI wearables that blend seamlessly into daily life...
A developer successfully built a sentiment analysis tool using local AI inference on a six-year-old laptop, proving on-device language models can solve real...
A new role is emerging across tech companies: the Associate Edge AI Engineer. These specialists deploy AI directly on devices, not in data centers, creating a...
AI models sized between 3 billion and 30 billion parameters are moving from cloud data centers to edge devices, enabling real-time decisions without internet...
A new M.2-compatible AI module packs 60 TOPS of processing power and 32GB memory into a slot most users leave empty, enabling 20-billion-parameter AI models on...
As AI moves from training to continuous decision-making, companies face a critical choice: keep intelligence centralized or distribute it locally.
Major open-source AI frameworks released critical updates in April 2026, making it practical to run capable AI models on everyday devices without cloud...
Smartwatches and earbuds are gaining dedicated neural processors for the first time, with flagship phones already delivering 100+ TOPS of AI compute.
Embedded AI is reshaping how devices process intelligence locally, not in the cloud.
Korean AI chip maker DEEPX is rapidly expanding into Japan with its energy-efficient DX-M1 neural processor, targeting the country's surging demand for...
Employees are quietly running AI models on laptops without IT approval, creating security risks that traditional cloud monitoring can't detect.
Leaked procurement documents reveal Chinese institutions are using coded language and technical specifications to secretly acquire Nvidia H200 chips despite...
A new framework reveals the real reason AI wearables end up in drawers: buyers focus on features instead of solving actual problems.
Qualcomm's leader argues the real AI winner won't dominate data centers but local devices. Here's why the market may be underestimating this shift.
Furiosa AI and Rebellion are establishing European footholds to capture sovereign AI demand with neural processing units offering higher power efficiency than...
AI coding agents perform better with established frameworks than writing code from scratch, despite higher token costs.
ST's Neural-ART architecture tackles edge AI's hidden bottleneck: data movement consumes half of system power.
Google released Eloquent, an on-device dictation app using its Gemma AI model, with no fanfare.
Edge AI processors are severely underutilized, wasting most of their computing power.
Huawei confirmed its first AI glasses with integrated camera launch for April 21, mirroring Ray-Ban Meta's feature set.
Apple's March 2026 event will introduce M5 chips with significantly upgraded NPUs, positioning local AI processing as central to its product strategy across...
Battery life in edge AI devices depends less on processor speed than on how computing tasks are distributed across multiple specialized chips, according to...
Edge AI is reshaping how industries handle data in 2026, with Microsoft, Google, and Advantech leading a shift toward local processing.
Computing is swinging back toward decentralization as neural processing units arrive in laptops.
AI accelerators demand radical new testing methods to catch failures across thousands of cores and complex multi-die packages.
AI-RAN transforms wireless networks into active computing layers for real-time factory operations, enabling millisecond-speed decisions that cloud computing...
Fujitsu is developing a 1.4-nanometer neural processing unit specialized for AI inference, promising lower power consumption than GPUs.
Google released LiteRT-LM, a production-grade framework for running large language models directly on edge devices without cloud dependency, signaling a major...
Meta's Ray-Ban Display AI glasses are unavailable in Europe due to strict battery regulations and AI compliance rules, despite being a hit in the US.
Consumer electronics will dominate the edge AI chip market through 2036, with neural processing units becoming standard in smartphones and PCs.
Makers and educators are combining tiny AI models with affordable hardware to solve real problems locally.
Puriosa AI's Renegade neural processor launches on Samsung SDS cloud in July, handling 7.4x more requests than Nvidia's RTX Pro 6000 with the same power...
The AI wearables market has consolidated dramatically in 2026. Devices that tried to replace smartphones failed, while those with focused use cases like task...
ASUS's new UGen300 USB AI Accelerator brings 40 AI TOPs of processing power to any device via a single USB port, enabling instant AI inference without cloud...
Flutter developers are building privacy-first AI apps that run machine learning models directly on phones, eliminating cloud dependencies.
