/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Source: Microsoft has completed training a family of models, codenamed MAI, and is experimenting with swapping out OpenAI's models in Copilot for the MAI models

Mustafa Suleyman wasn't getting the answers he wanted.  —  Last fall, during a video call with senior leaders at OpenAI and Microsoft

The Information Aaron Holmes

Context & Ripple Effects

Microsoft had already begun building MAI-1 as an in-house model effort and later explored adding third-party models to 365 Copilot amid cost and speed concerns. The reported completion of MAI turns that earlier in-house model program into a potential deployment option.

The move also fits Microsoft’s broader testing of alternatives from other labs for Copilot. Rather than treating a single supplier as the default, Microsoft is assembling leverage across internal and external model sources.

First-order effects

  • Microsoft gains a trained MAI family to evaluate directly against OpenAI models in Copilot, creating a credible internal fallback rather than relying solely on external options.
  • OpenAI faces a potential reduction in its role inside Copilot if Microsoft’s experiments translate into production use; the report does not establish that a replacement has occurred.

Second-order effects

  • Microsoft can use MAI results alongside its testing of models from other AI labs to compare capability, cost, and speed across suppliers, strengthening its negotiating position.
  • Other model providers have a clearer incentive to make their systems easier for large software platforms to evaluate and integrate, as Copilot becomes a contested distribution channel.

Third-order effects

  • If large application platforms routinely combine proprietary and outside models, frontier-model providers may increasingly compete for placement feature by feature rather than win exclusive platform-wide roles.
  • The pattern points toward AI procurement becoming a core product capability: buyers with distribution, compute, and model teams can exert more control over which models reach end users.

The trend: Major AI product platforms are shifting from dependence on one model partner toward multi-model portfolios designed to improve performance, economics, and bargaining power.

Discussion

  • @ecommerceshares @ecommerceshares on x
    Looks like $MSFT is abandoning the OpenAI ship.... Masayoshi Son top-ticked AGAIN. When will the Saudis say “This is enough! Stop fucking around, we want our cash back!”
  • @glinden Greg Linden on bluesky
    Yes, if open source or homegrown LLM models are almost as good and much cheaper to deploy, everyone is going to use the cheap models instead.  This is a big problem for OpenAI and other startups trying to charge a premium for models that aren't clearly better when actually used i…