Microsoft replaces OpenAI's image-generating models with its MAI models in its products; Mustafa Suleyman says MAI models are ~85% cheaper to run in PowerPoint
Microsoft Corp. is replacing OpenAI's image-generating models with its own technology in popular products like PowerPoint and Bing.
Context & Ripple Effects
Microsoft’s in-house model strategy has moved from experimentation to broader product substitution: it was previously reported to be replacing external models in Excel and Outlook, while a lower-cost efficient MAI image model had already been introduced for production use.
The shift also follows Microsoft’s earlier effort to reduce the cost of AI features in Bing through more efficient models. Extending MAI to high-volume consumer and productivity surfaces makes the cost-performance trade-off commercially consequential, not just a research result.
First-order effects
- Microsoft moves image-generation inference in PowerPoint and Bing from OpenAI to MAI, giving its own model family a larger share of deployed workload.
- At the reported roughly 85% lower PowerPoint running cost, Microsoft can reduce the operating cost of image features; OpenAI loses a portion of Microsoft product usage for this modality.
Second-order effects
- The deployment tests whether Microsoft can preserve product quality while treating model choice as a procurement decision based on cost and fit, reinforcing the earlier replacement of outside models in Office applications.
- OpenAI and other model suppliers face greater pressure to demonstrate differentiated performance or economics where major distributors can develop and deploy credible substitutes.
Third-order effects
- If similar substitutions spread across tasks, large AI distributors will gain more leverage over foundation-model suppliers by being able to route workloads among proprietary and external models.
- The industry may increasingly compete on cost per useful task and operational control, rather than on a single frontier model’s capabilities alone; the durability of that shift depends on MAI’s performance in real product use.
The trend: Microsoft’s MAI rollout is part of AI industrialization in which platform owners internalize selected model workloads to lower inference costs and strengthen supplier bargaining power.