All articles tagged with #edge ai
Google's Gemma 4 open models gain a 3x speed boost on local hardware through Multi-Token Prediction (MTP) drafters that guess future tokens and verify them in parallel with the main model, using shared memory and sparse decoding to accelerate generation without sacrificing quality. The approach works across devices from Pixel phones to Apple M4, with varying gains by hardware and under Apache 2.0 licensing.
Intel announced two leadership appointments to strengthen its core client computing business and advance its AI agenda: Alex Katouzian will become EVP and GM of the Client Computing & Physical AI Group, aligning client computing with edge AI, robotics and autonomous machines; Pushkar Ranade is appointed CTO, directing technology strategy across quantum computing, neuromorphic computing, photonics and novel materials. Both report to CEO Lip-Bu Tan, underscoring Intel’s push to accelerate AI-enabled computing across devices and future technologies.
Google’s Gemma 4 proves you don’t need a giant parameter count to punch above your weight: with 2.3 billion parameters it rivals 70B-scale models in performance while running offline on under 1.5 GB of RAM. The model is designed for edge use, features a 128K context window, and supports text, vision, and audio inputs in a multimodal architecture under an Apache 2.0 open-source license. It also handles 140 languages and emphasizes privacy and offline operation, making it suitable for resource-constrained devices. Notable caveats include coding task limitations and incomplete iOS bindings, with ongoing community-driven improvements anticipated.
Qualcomm unveiled the Snapdragon Wear Elite at MWC 2026, a wearable-focused chip with a Hexagon NPU for on-device edge AI and ultra-low‑power Micro-Power Wi‑Fi, delivering ~5x CPU performance and ~7x faster app launches, plus about 30% longer daily battery life and 50% charge in 10 minutes; Qualcomm envisions wearables—smartwatches, glasses, rings, and AI pins—becoming independent AI hubs in a distributed ecosystem, extending computing beyond smartphones.
TrendForce predicts a significant expansion in edge AI applications in 2024, with AI servers paving the way for AI PCs and other terminal devices. The global AI server market is projected to exceed 1.6 million units, growing at a rate of 40%, while AI PCs are expected to meet Microsoft’s benchmark of 40 TOPS in computational power. Qualcomm, AMD, and Intel are racing to meet these requirements, with LPDDR expected to represent a significant portion of PC DRAM demand due to the need for faster data transmission in AI PCs.
TrendForce predicts a significant expansion in edge AI applications in 2024, with AI servers and AI PCs playing a crucial role. The global AI server market is projected to exceed 1.6 million units, growing at a rate of 40%, while AI PCs are expected to meet Microsoft’s benchmark of 40 TOPS in computational power. Qualcomm, Intel, and AMD are set to intensify competition in the AI PC market, with LPDDR expected to represent approximately 30–35% of PC DRAM demand this year.