Microsoft Replaces OpenAI Image Model in PowerPoint and Bing with In-House AI
Microsoft is replacing the OpenAI image model in PowerPoint and Bing with its own AI technology.
The company claims this move brings faster speeds, lower costs, and higher quality results.
The replacement reduces external dependencies, directing funds towards internal model development, benefiting Microsoft through cost control, while OpenAI faces pressure in related application scenarios.
Source: Public Information
ABAB AI Insight
Microsoft has had a long-term deep collaboration with OpenAI, but in recent years has accelerated the development of its own MAI series models, covering images, speech, etc., and prioritizing the deployment of internal solutions in products like Office and Bing to achieve cost and performance optimization.
Its capital path reduces token and inference costs through large-scale in-house development, motivated by the desire to shift routine high-frequency tasks from general cutting-edge models to specialized efficient models, enhancing profit margins and reducing reliance on a single partner.
Similar cases can be seen with Google prioritizing in-house models in search, or Amazon replacing some external computing power with its own chips; currently, major companies are in a transitional phase from dependency on collaboration to a balanced hybrid of in-house development.
Essentially, this is a technological substitution: in-house specialized models are replacing external general models as the default option for products, driven by the stringent cost and latency requirements of high-frequency scenarios, making internal optimization more economical and reshaping AI supply chain relationships.
ABAB News · Cognitive Laws
- High-frequency tasks prioritize in-house models
- Cost advantages determine default choices
- Collaboration and in-house development in parallel become the new norm.