Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs
Microsoft has introduced its third-generation in-house AI chip, the Maia 300, designed to act as the primary AI inference system for its internal products and services. The technology aims to reduce the company's heavy dependence on Nvidia and AMD GPUs while curbing cloud computing expenses. Microsoft plans to deploy the Maia chips at gigawatt scale starting next year, with full production rollouts set for 2027. The company claims the hardware will deliver 40 percent more tokens per megawatt than leading GPU-based systems, enhancing the cost efficiency of running applications like Copilot.
Listen to Live Briefing
Real-time synthesized voice briefing · Live Feeds Desk
- ✓ Microsoft revealed the Maia 300, its third-generation in-house AI chip, to serve as the primary AI inference system for its products and services.
- ✓ Microsoft plans to deploy its Maia 300 AI chips at gigawatt scale starting next year, with production rollouts planned for 2027.
- ✓ The Maia 300 chip is designed to deliver 40 percent more tokens per megawatt than leading GPU-based systems.
What changed
Microsoft officially announced the Maia 300 AI accelerator and outlined plans for multi-gigawatt deployments.
Live updates
-
Microsoft Unveils Maia 300 AI Chip to Curb Nvidia Reliance
Microsoft has introduced its third-generation in-house AI chip, the Maia 300, designed to act as the primary AI inference system for its internal products and services. The technology aims to reduce the company's heavy dependence on Nvidia and AMD GPUs while curbing cloud computing expenses. Microsoft plans to deploy the Maia chips at gigawatt scale starting next year, with full production rollouts set for 2027. The company claims the hardware will deliver 40 percent more tokens per megawatt than leading GPU-based systems, enhancing the cost efficiency of running applications like Copilot.
Why it matters
Major technology companies are increasingly designing proprietary silicon to escape supply bottlenecks and lower the massive financial costs associated with running artificial intelligence workloads. Microsoft has relied heavily on external suppliers like Nvidia to power its cloud infrastructure and AI offerings. Developing custom accelerators allows the company to optimize performance specifically for its software ecosystem while retaining better control over hardware margins.
What is confirmed
- Microsoft revealed the Maia 300, its third-generation in-house AI chip, to serve as the primary AI inference system for its products and services.
- Microsoft plans to deploy its Maia 300 AI chips at gigawatt scale starting next year, with production rollouts planned for 2027.
- The Maia 300 chip is designed to deliver 40 percent more tokens per megawatt than leading GPU-based systems.
Still unconfirmed
- The Maia 300 chip might make Teams, Copilot, and Cowork much better to use.
What to watch next
- Actual deployment volumes of the Maia 300 in Microsoft data centers next year
- Performance benchmarks comparing Maia 300 directly against competing Nvidia hardware
confidence 95%Sources used for this update (25)
- account.microsoft.com — Microsoft account | Sign In or Create Your Account Today ...
- en.wikipedia.org — Microsoft - Wikipedia
- CNBC — Microsoft to rely more on homegrown AI chips, reducing dependence on Nvidia
- Yahoo Finance — Microsoft reveals Maia 300 AI chip as it works to reduce dependence on Nvidia, AMD
- The Official Microsoft Blog — Microsoft to expand Azure Maia system to multi-gigawatt deployments
- Reuters — Microsoft plans huge deployment of its own AI chips, with an eye on saving costs
- The Verge — Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs
- Microsoft Azure — Deep dive into Azure Maia 300 architecture
- Tech Xplore — Microsoft bets on homegrown AI chip to challenge Nvidia
- Moomoo — U.S. Stock Market Movers | Reports Suggest Microsoft (MSFT.US) to Deploy Self-Developed AI Chips on a Large Scale Next Year; Shares Rise Nearly 1% in Early Trading
- Bloomberg.com — Microsoft Prepares to Put Gigawatts of Its Chips in Data Centers
- bsky.app — @starlink85.bsky.social
Community Sentiment: How do you assess this situation?
Voice your perspective · Real-time aggregated sentiment from the Live Feeds community