.jpg?w=1024)
Microsoft is discovering that one way to reduce the cost of AI is to use less technology from its biggest AI partner.
The company is replacing OpenAI image models with its in-house Microsoft AI technology across features in PowerPoint, Bing, and OneDrive. Microsoft says the change has made some services faster while cutting PowerPoint image-generation costs by roughly 85%.
The rollout is part of a much broader shift: Microsoft AI CEO Mustafa Suleyman says the company’s MAI models are already deployed in more than half of Microsoft’s products, further reducing its dependence on outside model developers.
"It's faster, it's cheaper, it's higher quality, it drives better retention," Suleyman said in an interview, according to Bloomberg.
New models spread across Microsoft's ecosystem
Microsoft said Bing Image Creator now runs entirely on its MAI-Image-2.5 model. The same technology has also become the default model for several image-editing features in OneDrive, where the company reported faster performance, greater efficiency and more users saving AI-generated images.
Beyond image generation, Microsoft also introduced MAI-Voice-2-Flash, a voice model designed for call centers and voice agents. The company said customers including T-Mobile and easyJet are already using the technology through Dynamics 365 Contact Center and Azure.
Microsoft has also been expanding MAI into text-generation workloads. Bloomberg previously reported that the company had begun replacing OpenAI and Anthropic models inside Office applications. Microsoft says one MAI model used in Excel performs on par with OpenAI's GPT-5.6 for common tasks while operating at a lower cost.
Suleyman, per Bloomberg, noted that MAI models are now deployed in more than half of Microsoft's products and are being tested across the rest of its software portfolio.
"It is a very strong showing, and we're going to continue to push extremely hard," Suleyman said.
Partnership with OpenAI remains, but priorities are changing
Microsoft's shift does not end its relationship with OpenAI. Under their long-standing partnership, Microsoft continues to receive access to OpenAI's models. However, Microsoft is responsible for supplying the computing infrastructure required to run those models, making efficiency a growing priority.
The company has invested roughly $13 billion in OpenAI, but the partnership has become less exclusive as OpenAI expands to additional cloud providers while Microsoft increasingly develops its own foundation models.
Microsoft has also said frontier models from OpenAI and Anthropic will continue to be used for more demanding AI tasks. In contrast, its in-house MAI models handle workloads for which similar performance can be achieved at a lower cost.
What this signals for Microsoft's AI strategy
Microsoft's latest move highlights that the next stage of the AI race is no longer just about building the most capable model; it is also about operating AI economically at massive scale. For products used by hundreds of millions of people, even small efficiency gains can translate into substantial infrastructure savings.
For customers, the transition could mean faster AI features and more responsive image-generation tools without noticeable changes to how they use Microsoft applications. The risk is that Microsoft now carries greater responsibility for proving its own models can consistently match or exceed the quality users have come to expect from OpenAI across a wider range of creative tasks.
Also Read: How to get better results from AI image generators with these 8 prompting tricks.