Microsoft to Launch Maia 300 AI Accelerator in 2027 with Enhanced Token Efficiency
- Published
- Source
- InterestingEngineering.com
Microsoft has announced plans to deploy its next-generation Maia 300 AI accelerator in 2027, aiming for a 40% increase in tokens per megawatt compared to current GPU systems. This new chip will feature a 96-accelerator rack configuration, delivering up to seven times the token throughput of its predecessor, Maia 200, which is already in production use.
The Maia 300 is part of Microsoft's strategy to develop custom silicon for large-scale AI inference, addressing the growing demands of memory-intensive workloads and complex AI tasks. With a system-level architecture that includes eight HBM4 memory stacks and 18 terabytes per second of peak bandwidth, the Maia 300 is designed to optimize performance while operating within a 950-watt thermal design power envelope.
Looking ahead, Microsoft expects its GitHub Copilot and Microsoft Foundry to adopt the Maia 300 by the end of 2026. The company has not disclosed independent benchmarks for the Maia 300 but claims it will significantly outperform existing GPU-based systems in terms of token efficiency and throughput.
Original report
Microsoft’s new Maia 300 packs 96 chips into one 167-kW rack with 40% token efficiency
InterestingEngineering.com · Aamir KhollamThis briefing is an independently written summary based on publicly available reporting and is provided for industry information and news discovery. The original report and source publication are credited and linked where applicable. RobotToday does not claim ownership of third-party source material.
Rights concerns? Contact [email protected] with the relevant URL and details. We will review the matter and take appropriate action where warranted.
More related news
- 1
- 2
- 3
- 4
- 5
- 6
- 7
-
8
RobotToday Weekly October 5 – 9, 2026 RobotToday · 2 days ago
- 9
- 10
- 11
- 12
Related Suppliers
Microsoft Corporation (Robotics)
United States
Microsoft provides robotics developers ROS-on-Windows support, the AirSim open-source simulation platform, and Azure cloud services for robot fleets and AI.
Mikron Automation
Switzerland
Swiss builder of automated assembly and test systems, mainly for auto-injectors, pens and inhalers, with 4,500+ systems installed; part of SIX-listed Mikron.
Hugging Face
United States
Open-source AI platform and robotics enabler; home of LeRobot, 2M+ models, and the Reachy Mini desktop robot.
Miru
United States
US robotics config management SaaS (YC S24, est. 2024); OTA deployment and fleet-scale schema versioning for robots and IoT.
Anthropic
United States
AI safety and research company and public benefit corporation that builds the Claude family of large language models.
Fireworks AI
United States
Redwood City AI infra startup (2022) serving open-source LLMs like Kimi K3 via API, used by Cursor, Vercel and Notion.
Zyphra Technologies Inc.
United States
Palo Alto AI company building compact language, vision-language, and text-to-speech models including the open-source Zamba and Zonos model families.
Nivavi
Netherlands
Zwolle startup whose Bart kit connects to a tractor's CAN bus and RTK-GPS to make existing tractors drive and work autonomously.
Cognida.ai
United States
Cognida.ai is an enterprise AI consulting and engineering firm whose Zuno platform includes computer-vision tooling for industrial image analysis.
d-Matrix
United States
Fabless semiconductor startup whose Corsair platform uses digital in-memory compute to accelerate generative AI inference in datacenters.
LG Electronics Inc.
Korea, Republic of (South)
Seoul electronics group (est. 1958, KRX: 066570) selling CLOi GuideBot, ServeBot and CarryBot service robots; controls Bear Robotics (51% since 2025).
Multi Industries of Greenville
United States
US automation integrator founded 1997 in Taylors, SC; serves automotive and industrial manufacturers with mechanical, electrical, and commissioning services.
Robotics needs a service framework.
RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.
Join the RobotToday community on LinkedIn
Daily robotics news, in-depth analysis, conference highlights, and discussions with professionals worldwide.
