{"slug": "steel-revolutionizing-energy-efficient-ai-on-laptops", "title": "STEEL: Revolutionizing Energy-Efficient AI on Laptops", "summary": "Researchers introduced STEEL, the first open-source implementation of FlashAttention optimized for NPUs, achieving up to 22.8x speedup on AMD's Ryzen AI 9 HX 370 SoC. The approach reduces energy use by 9.17x over CPUs and 1.75x over GPUs, enabling efficient on-device AI without cloud offloading.", "body_md": "# STEEL: Revolutionizing Energy-Efficient AI on Laptops\n\nSTEEL introduces a groundbreaking approach to running large language models on NPUs in laptops, achieving significant energy and speed advantages.\n\nAs large language models become integral to our daily tech interactions, the need for energy-efficient solutions is more pressing than ever. The rising trend of integrating these models in operating system workflows demands innovation, especially on laptop-class systems-on-chip (SoCs).\n\n## Breaking New Ground with STEEL\n\nSTEEL emerges as a pioneering solution, marking the first open-source implementation of FlashAttention optimized for XDNA-like neural processing units (NPUs). The paper's key contribution: a novel dataflow formulation of prefill [attention](/glossary/attention) that capitalizes on spatial parallelism and on-chip memory. This enables a leap in efficiency.\n\nWhy does this matter? Cloud offloading, while popular, brings unwanted reliability and privacy issues. By shifting the heavy lifting to NPUs, STEEL addresses these concerns head-on. But, mapping attention mechanisms to NPUs isn't straightforward, owing to diverse architectures and explicit programming models.\n\n## Performance and Efficiency Gains\n\nOn the performance front, STEEL doesn't disappoint. Tested on AMD's Ryzen AI 9 HX 370 SoC, it dwarfs traditional CPU and [GPU](/glossary/gpu) implementations, slashing energy use by 9.17x and 1.75x, respectively. That's no small feat.\n\nBut the real showstopper? On XDNA 1, STEEL reduces latency by an average of 9.6x over prior SOTA, while on XDNA 2, it achieves a staggering 22.8x speedup compared to layer-by-layer attention implementations.\n\n## Implications for the Future\n\nThis development raises a important question: Are NPUs the future of on-device AI processing? STEEL's success suggests we're heading in that direction. The ablation study reveals that optimizing for architectural specifics can lead to substantial gains in both speed and energy efficiency.\n\nWhat's missing, though, is broader adoption across varying device types. For STEEL, or similar frameworks, to become a staple, more devices need to embrace these tailored NPU architectures. Will manufacturers take the bait? The potential benefits are clear.\n\nCode and data are available at the project's repository, offering a transparent and reproducible pathway for others to follow or build upon. This builds on prior work from the field but takes it a step further with practical, demonstrable advancements.\n\nGet AI news in your inbox\n\nDaily digest of what matters in AI.", "url": "https://wpnews.pro/news/steel-revolutionizing-energy-efficient-ai-on-laptops", "canonical_source": "https://www.machinebrief.com/news/steel-revolutionizing-energy-efficient-ai-on-laptops-rp2v", "published_at": "2026-07-13 10:24:02+00:00", "updated_at": "2026-07-13 10:49:59.626399+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research", "ai-infrastructure"], "entities": ["STEEL", "AMD", "Ryzen AI 9 HX 370", "XDNA", "FlashAttention"], "alternates": {"html": "https://wpnews.pro/news/steel-revolutionizing-energy-efficient-ai-on-laptops", "markdown": "https://wpnews.pro/news/steel-revolutionizing-energy-efficient-ai-on-laptops.md", "text": "https://wpnews.pro/news/steel-revolutionizing-energy-efficient-ai-on-laptops.txt", "jsonld": "https://wpnews.pro/news/steel-revolutionizing-energy-efficient-ai-on-laptops.jsonld"}}