Microsoft plans to scale its own Maia AI chips to gigawatt levels by 2027
Source: The Decoder (opens in a new tab) · Matthias Bastian
Intel Summary
Microsoft plans to make its custom Azure Maia silicon its primary AI inference hardware across company products beginning in 2027, deploying at multi-gigawatt scale. According to the company, the upcoming Maia 300 is designed to deliver 40 percent more tokens per megawatt compared to leading GPU systems.
Why It Matters
Transitioning core inference workloads to proprietary silicon aims to improve power efficiency per token and reduce dependency on external GPU suppliers as hyperscale datacenter energy consumption accelerates.
Organizations & Entities
Topics
Related Intelligence
- ReportAlso involving Microsoft
Amazon drops data center NDAs, and AI agents want your credit card
Amazon announced it will stop requiring non-disclosure agreements when negotiating data center developments with local governments, following a similar move by Microsoft. The shift follows growing public opposition and hundreds of proposed or enacted municipal moratoriums on AI infrastructure projects from New York to San Francisco. The report also highlights a growing cohort of startups building consumer AI agents designed to access payment credentials.
TechCrunch - ReportAlso involving Microsoft
Why the SaaSpocalypse Isn’t Over For Workday
Enterprise customers are deploying AI agents from providers such as Anthropic and Microsoft to pull data from Workday for HR and financial analysis, according to The Information. This pattern allows organizations to query records without users directly visiting the application interface, placing constraints on growth for traditional SaaS platforms.
The Information - DevelopmentNewAlso involving Microsoft
Meta and Microsoft scale back internal usage of Anthropic Claude models
Meta and Microsoft are significantly scaling back internal usage of Anthropic's Claude models to promote their own AI tools. Microsoft reduced its cloud division's monthly per-employee Claude budget from $100,000 to $10,000, while Meta decreased its Claude Code user base by half to 30,000 seats. Claims are as reported; this summary makes no determination about accuracy or significance.
1 reporting source - ReportAlso involving Microsoft
Microsoft gives Copilot another makeover, adding an Autopilot agent and usage-based billing
Microsoft is restructuring its Copilot application into three sections—Home, Code, and an autonomous cloud agent named Autopilot built on OpenClaw. According to Microsoft, Autopilot operates continuously in the cloud to monitor Teams channels and execute tasks independently. The company is also shifting pricing for Autopilot and Code from flat-rate subscriptions to usage-based billing.
The Decoder