{ "version": 1, "event_id": "evt_26e5b0cda31c1ae7", "url": "https://xiyu.news/events/evt_26e5b0cda31c1ae7/", "json": "https://xiyu.news/api/events/evt_26e5b0cda31c1ae7.json", "type": "other", "status": "monitoring", "category": "technology", "title": { "zh": "微软新AI芯片降低对英伟达GPU的依赖", "en": "Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs" }, "current_state": { "zh": "微软发布第三代Maia 300 AI芯片，定位为自家产品和服务的主要推理系统，计划从2027年起扩大至数吉瓦级部署，以减少对英伟达GPU的依赖；模型训练继续使用英伟达和AMD。新增细节：明年可能部署超过100万颗Maia 300；计算芯粒采用台积电2nm工艺，每颗288GB HBM4显存，带宽18TB/s；单机柜容纳96颗芯片，功率约167kW，推理吞吐量最高为Maia 200的7倍，每兆瓦Token数比领先GPU系统多约40%；上一代Maia 200已在Microsoft Copilot中运行。", "en": "Microsoft announced its third-generation Maia 300 AI chip, positioned as the primary inference system for its own products and services, with plans to scale to multi-gigawatt deployments from 2027 to reduce reliance on Nvidia GPUs; model training continues on Nvidia and AMD. New details: over 1 million Maia 300 units may be deployed next year; compute dies use TSMC 2nm process, each with 288GB HBM4 memory and 18TB/s bandwidth; each rack holds 96 chips at about 167kW, with inference throughput up to 7x Maia 200 and ~40% more tokens per MW than leading GPU systems; previous-generation Maia 200 is already running in Microsoft Copilot." }, "first_seen_at": "2026-10-12T16:21:16.038890+00:00", "last_updated_at": "2026-10-13T03:47:30.490841+00:00", "last_material_change_at": "2026-10-13T03:47:30.490841+00:00", "confidence": 0.75, "updates_count": 3, "sources_count": 3, "entities": [ "gpus", "maia", "maia-300", "microsoft", "nvidia" ], "identifiers": [], "topics": [ "ai-chips", "ai-infrastructure", "data-centers", "inference", "maia-300", "microsoft", "nvidia" ], "updates": [ { "update_id": "upd_fa9bc408c6e96277", "event_id": "evt_26e5b0cda31c1ae7", "occurred_at": "2026-10-12T13:00:00Z", "published_at": "2026-10-12T13:00:00Z", "first_seen_at": "2026-10-12T16:21:16.038890Z", "time_precision": "published", "update_type": "initial", "material_change": true, "title_zh": "微软新AI芯片降低对英伟达GPU的依赖", "title_en": "Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs", "what_changed_zh": "微软发布第三代Maia 300 AI芯片，定位为自家产品和服务的主要推理系统，具备更优的成本效益，并支持推理和智能体工作负载，从而减少对英伟达GPU的依赖。", "what_changed_en": "Microsoft announced its third-generation Maia 300 AI chip, positioned as the primary inference system for its own products and services with improved cost efficiency and support for reasoning and agentic workloads, reducing its reliance on Nvidia GPUs.", "current_state_zh": "微软发布第三代Maia 300 AI芯片，定位为自家产品和服务的主要推理系统，具备更优的成本效益，并支持推理和智能体工作负载，从而减少对英伟达GPU的依赖。", "current_state_en": "Microsoft announced its third-generation Maia 300 AI chip, positioned as the primary inference system for its own products and services with improved cost efficiency and support for reasoning and agentic workloads, reducing its reliance on Nvidia GPUs.", "detailed_summary_zh": "Microsoft announced its third-generation Maia 300 AI chip, positioned as the primary inference system for its own products and services with improved cost efficiency and support for reasoning and agentic workloads, reducing its reliance on Nvidia GPUs.", "detailed_summary_en": "Microsoft announced its third-generation Maia 300 AI chip, positioned as the primary inference system for its own products and services with improved cost efficiency and support for reasoning and agentic workloads, reducing its reliance on Nvidia GPUs.", "background_zh": "", "background_en": "", "community_discussion_zh": "", "community_discussion_en": "", "market_impact_zh": "", "market_impact_en": "", "importance_score": 7.0, "references": [], "confidence": 0.75, "story_ids": [ "rss:www.theverge.com_rss_ai-artificial-intelligence_index.xml:dd7e55f44c038cc1" ], "sources": [ { "url": "https://www.theverge.com/news/1009465/microsoft-maia-300-chip-improvements-rollout", "label": "The Verge AI", "source_type": "rss", "official": false } ] }, { "update_id": "upd_6a53ef33b9ef4674", "event_id": "evt_26e5b0cda31c1ae7", "occurred_at": "2026-10-12T13:04:02Z", "published_at": "2026-10-12T13:04:02Z", "first_seen_at": "2026-10-12T16:20:55.196602Z", "time_precision": "published", "update_type": "confirmation", "material_change": true, "title_zh": "微软将加大自研AI芯片使用，Maia 300预计2027年投产", "title_en": "微软将加大自研AI芯片使用，Maia 300预计2027年投产", "what_changed_zh": "新增细节：Maia 300预计2027年投产；将在Maia系列上部署数吉瓦级算力；每个数据中心机架可搭载96颗Maia 300芯片，是上一代Maia 200的两倍；与GPU相比，每机架AI Token处理量最高可达7倍，每兆瓦Token数量提升约40%；Evercore ISI预计到2027年Q1 Maia芯片将占微软AI芯片支出约10%；模型训练仍使用英伟达和AMD芯片；目前不向其他企业开放租用。", "what_changed_en": "New details: Maia 300 expected to enter production in 2027; multi-gigawatt deployment on Maia series; each data center rack can hold 96 Maia 300 chips, double the previous Maia 200; up to 7x AI token throughput per rack vs GPU and ~40% more tokens per MW; Evercore ISI estimates Maia chips will account for ~10% of Microsoft AI chip spending by Q1 2027; model training still uses Nvidia and AMD chips; not yet available for rental by other enterprises.", "current_state_zh": "微软发布第三代Maia 300 AI芯片，定位为自家产品和服务的主要推理系统，具备更优的成本效益，并支持推理和智能体工作负载，从而减少对英伟达GPU的依赖；目标于2027年投产，并计划进行数吉瓦级的内部部署，模型训练则继续使用英伟达和AMD。", "current_state_en": "Microsoft announced its third-generation Maia 300 AI chip, positioned as the primary inference system for its own products and services with improved cost efficiency and support for reasoning and agentic workloads, reducing its reliance on Nvidia GPUs; production targeted for 2027 and multi-gigawatt internal deployment planned, with model training continuing on Nvidia and AMD.", "detailed_summary_zh": "Microsoft unveiled its third-generation in-house AI accelerator, Maia 300, targeting 2027 production and multi-gigawatt internal deployment to cut reliance on Nvidia GPUs for inference while keeping Nvidia and AMD silicon for model training.", "detailed_summary_en": "Microsoft unveiled its third-generation in-house AI accelerator, Maia 300, targeting 2027 production and multi-gigawatt internal deployment to cut reliance on Nvidia GPUs for inference while keeping Nvidia and AMD silicon for model training.", "background_zh": "", "background_en": "", "community_discussion_zh": "", "community_discussion_en": "", "market_impact_zh": "", "market_impact_en": "", "importance_score": 7.0, "references": [], "confidence": 0.95, "story_ids": [ "telegram:theblockbeats:200078" ], "sources": [ { "url": "https://m.theblockbeats.info/flash/371411?from=telegram", "label": "theblockbeats", "source_type": "telegram", "official": false } ] }, { "update_id": "upd_9173399a00e7a516", "event_id": "evt_26e5b0cda31c1ae7", "occurred_at": "2026-10-13T03:05:02Z", "published_at": "2026-10-13T03:05:02Z", "first_seen_at": "2026-10-13T03:47:30.490841Z", "time_precision": "published", "update_type": "confirmation", "material_change": true, "title_zh": "微软要让自研芯片成为AI推理主力，Maia 300明年或部署超百万颗", "title_en": "微软要让自研芯片成为AI推理主力，Maia 300明年或部署超百万颗", "what_changed_zh": "新增细节：微软明年可能部署超过100万颗Maia 300；Maia 300计算芯粒采用台积电2nm工艺，每颗配备288GB HBM4显存，内存带宽达18TB/s；单个机柜功率约167kW；上一代Maia 200已在Microsoft Copilot中运行，支持OpenAI、微软自家及开源模型。", "what_changed_en": "New details: Microsoft may deploy over 1 million Maia 300 units next year; Maia 300 compute dies use TSMC 2nm process, each with 288GB HBM4 memory and 18TB/s bandwidth; each rack draws about 167kW; previous-generation Maia 200 is already running in Microsoft Copilot, supporting OpenAI, Microsoft and open-source models.", "current_state_zh": "微软发布第三代Maia 300 AI芯片，定位为自家产品和服务的主要推理系统，计划从2027年起扩大至数吉瓦级部署，以减少对英伟达GPU的依赖；模型训练继续使用英伟达和AMD。新增细节：明年可能部署超过100万颗Maia 300；计算芯粒采用台积电2nm工艺，每颗288GB HBM4显存，带宽18TB/s；单机柜容纳96颗芯片，功率约167kW，推理吞吐量最高为Maia 200的7倍，每兆瓦Token数比领先GPU系统多约40%；上一代Maia 200已在Microsoft Copilot中运行。", "current_state_en": "Microsoft announced its third-generation Maia 300 AI chip, positioned as the primary inference system for its own products and services, with plans to scale to multi-gigawatt deployments from 2027 to reduce reliance on Nvidia GPUs; model training continues on Nvidia and AMD. New details: over 1 million Maia 300 units may be deployed next year; compute dies use TSMC 2nm process, each with 288GB HBM4 memory and 18TB/s bandwidth; each rack holds 96 chips at about 167kW, with inference throughput up to 7x Maia 200 and ~40% more tokens per MW than leading GPU systems; previous-generation Maia 200 is already running in Microsoft Copilot.", "detailed_summary_zh": "Microsoft announced it will scale its in-house Maia chips to multi-gigawatt deployments from 2027 to become the main inference platform for its AI products, with over one million Maia 300 units possibly deployed next year and specs of 2nm compute dies, 288GB HBM4, 18TB/s bandwidth and 96 chips per rack.", "detailed_summary_en": "Microsoft announced it will scale its in-house Maia chips to multi-gigawatt deployments from 2027 to become the main inference platform for its AI products, with over one million Maia 300 units possibly deployed next year and specs of 2nm compute dies, 288GB HBM4, 18TB/s bandwidth and 96 chips per rack.", "background_zh": "", "background_en": "", "community_discussion_zh": "", "community_discussion_en": "", "market_impact_zh": "", "market_impact_en": "", "importance_score": 7.5, "references": [], "confidence": 0.92, "story_ids": [ "telegram:theblockbeats:200166" ], "sources": [ { "url": "https://m.theblockbeats.info/flash/371498?from=telegram", "label": "theblockbeats", "source_type": "telegram", "official": false } ] } ] }