Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs
https://www.theverge.com/news/1009465/microsoft-maia-300-chip-improvements-rollout📌 微軟 Maia 300：想甩開對 Nvidia GPU 的依賴
TL;DR：微軟發布第三代自研 AI 晶片 Maia 300，宣稱效能大幅提升，2027 年起成為自家產品的主力推理系統。
當全世界的雲端巨頭都在排隊搶購 Nvidia GPU 時，微軟選擇走另一條路：自己做晶片，而且宣布這次要讓它成為公司 AI 服務的「主力」——不再只是備援方案。
🤔 第三代自研晶片，目標是「主力推理系統」
微軟今日發布第三代自研 AI 晶片 Maia 300，定位為「微軟產品與服務的主力 AI 推理系統」。它是去年一月才發布的 Maia 200 的後繼產品，而 Maia 200 目前已經在 Microsoft Copilot、GitHub Copilot、Microsoft Foundry 以及微軟自家 AI 模型的生產工作負載上運行。Azure 雲端部門執行副總裁 Scott Guthrie 表示，Maia 200 是公司機隊中，以實際 AI 服務指標衡量表現最好的基礎設施。
🧩 用 chiplet 架構換效能
Maia 300 採用新的 chiplet 架構，建構於 TSMC N2P 製程節點，將運算、記憶體控制器與 I/O 功能拆分開來。Azure 硬體與基礎設施團隊表示，這樣的設計讓更多矽晶面積能分配給運算效能，同時讓每個功能都能使用最適合自己的技術。
📊 宣稱 40% 更高能效，7 倍機架輸出量
Guthrie 表示，相較於主流 GPU 系統，Maia 300 預期能帶來「每百萬瓦 40% 更多 token 輸出量」；在 96 加速器的完整機架配置下，更可達到 Maia 200 的 7 倍機架層級 token 輸出量。微軟計畫從 2027 年開始，把 Azure Maia 家族擴展至數十億瓦等級的部署規模。
💡 為全端自研鋪路
Maia 300 的設計目標明確指向目前 LLM 工作負載的新趨勢：長上下文推理、延遲敏感的對話體驗，以及持續多步驟的 agentic 任務。微軟 AI 部門 CEO Mustafa Suleyman 表示，「我們必須從全端一路做到矽晶片層級」，並強調在 Maia 上運行微軟自家 AI 模型，是團隊的關鍵焦點——這也被視為微軟追趕 Google DeepMind、OpenAI、Anthropic 的策略之一。
⚠️ 目前仍是公司自己的說法
Maia 300 要到 2027 年才開始大規模投入生產工作負載，報導中引用的效能提升數字（40% 能效、7 倍輸出量）皆來自微軟自身說法，尚未見第三方獨立驗證；報導本身也未揭露具體的晶片微架構細節（如核心數、記憶體頻寬規格），對工程師而言目前還難以據此做出技術選型判斷。
🎯 實務啟示
如果你的團隊已經在 Azure 上跑推理工作負載，可以留意 Maia 300 於 2027 年量產後，是否會開放給一般客戶使用、以及屆時的定價與 API 相容性；在那之前，這則消息更適合當作觀察雲端廠商晶片自研趨勢的參考，而非立即影響架構決策的依據。
🔗 來源
- 標題：Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs
- 作者／機構：Tom Warren（The Verge）
- 連結：https://www.theverge.com/news/1009465/microsoft-maia-300-chip-improvements-rollout
#Microsoft #Maia300 #AIChip #Azure #Nvidia #AIInfrastructure #Chiplet #CloudComputing #Semiconductors #AIInference
原始資料 The Verge AI · 收集於 2026-10-12
摘要原文
News AI Gadgets Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs The new Maia 300 chip will start rolling out to production AI workloads in 2027. The new Maia 300 chip will start rolling out to production AI workloads in 2027. by Tom Warren Oct 12, 2026, 1:00 PM UTC Microsoft’s Maia 300 chip. Image: Microsoft Tom Warren is a senior correspondent and author of Notepad , who has been covering all things Microsoft, PC, and tech for over 20 years. Microsoft is announcing its third-generation Maia in-house AI chip today, designed to be the “primary AI inference system” for Microsoft’s products and services. Maia 300 not only improves the cost and efficiency of running AI workloads in the cloud, it’s also more capable of handling the latest reasoning models and agentic tasks. Maia 300 will start being rolled out to production in 2027, and it has been designed with efficiency in mind. “We expect Maia 300 to deliver 40 percent more tokens per megawatt compared to leading GPU-based systems,” says Scott Guthrie , executive vice president of Microsoft’s cloud group. “And in a full 96-accelerator configuration, it delivers up to 7x the rack-scale token throughput of Maia 200.” These efficiency improvements will allow Microsoft to rely mainly on its in-house AI chips soon, instead of ones from Nvidia and AMD. “We will scale our Azure Maia family to multi-gigawatt deployments starting in 2027, making Maia the primary AI inference system for Microsoft products and services,” says Guthrie. Maia 300 chips will start being used for production workloads in 2027. Image: Microsoft Maia 300 is a successor to the 200, which was only announced in January . Maia 200 is already running on production AI workloads across Microsoft, including Microsoft Copilot, GitHub Copilot, Microsoft Foundry, and Microsoft’s own AI models. “Maia 200 is the best performing AI infrastructure in our fleet, measured in real world metrics for AI services,” says Guthrie. “Maia 300 is a complete AI system designed for the new types of AI workloads emerging today: Long-context reasoning, latency-sensitive conversational experiences, and sustained, multi-step agentic tasks.” Microsoft is using a new chiplet architecture to get its all-important efficiency gains. Built on TSMC’s N2P process node, the latest Maia 300 architecture separates compute, memory controller, and I/O functions. “This increases the silicon area dedicated to compute performance and allows each function to use the technology best suited to its role,” says the Azure hardware and infrastructure team. Maia 300 is also a key part of Microsoft’s own AI model work, and could help the company achieve its goal of catching up with Google DeepMind, OpenAI, and Anthropic. “We have to work across the full stack, all the way down to the silicon,” says Microsoft AI CEO Mustafa Suleyman. “At that scale we need a platform that we own, and that we co-design with our models. That’s why running our Microsoft AI models on top of the Maia chip is a critical focus for our team.” Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates. Tom Warren AI Chips Gadgets Microsoft News Tech
由 claude-code/sonnet 自動生成