What is the story about?
As the cost of building and running artificial intelligence services continues to climb, Microsoft is reportedly adjusting how it powers some of its most popular productivity software. According to a Bloomberg report, the company has begun using its internally developed AI models for a portion of requests in applications such as Word and Excel, reducing its reliance on external models from OpenAI and Anthropic for certain workloads.
The reported shift does not signal an end to Microsoft's partnerships with either AI company. Instead, it appears to be part of a broader strategy to balance performance with operating costs by assigning selected tasks to its own models while continuing to use third-party systems where appropriate.
The development comes at a time when major technology companies are reassessing the financial demands of generative AI, even as investment in the technology continues at pace.
Microsoft has spent the past year building a growing portfolio of proprietary AI models under its MAI branding. At its annual Build developer conference last month, the company introduced seven new MAI models, including an AI coding assistant and a text-to-image generation model, signalling its ambition to compete more directly in the foundation model space.
Bloomberg reported that some user prompts in Microsoft 365 applications are now being handled by these in-house models rather than exclusively by OpenAI or Anthropic. Previously, Microsoft had highlighted that many AI-powered features in its productivity suite were driven by models from those external partners.
When asked about the reported changes, Microsoft told TechCrunch that it had nothing further to add.
Developing proprietary models gives Microsoft greater flexibility over how AI services are deployed and may also help the company reduce the costs associated with running high-volume enterprise workloads.
Microsoft's reported approach reflects a wider trend across the technology sector, where companies are increasingly looking for ways to make AI services more financially sustainable.
After a period of aggressive spending on AI infrastructure and model deployment earlier this year, several large technology firms have reportedly started introducing measures aimed at controlling expenses. Companies including Amazon, Uber, Meta and Accenture have all been linked to efforts to optimise AI-related spending.
The economics of generative AI remain one of the industry's biggest challenges. Training advanced models requires enormous computing resources, while serving millions of user requests every day continues to generate significant infrastructure costs.
The pressure to reduce those expenses has also led some companies to explore lower-cost AI alternatives, including models developed in China. While these options may offer financial advantages, organisations continue to weigh them against concerns surrounding security, compliance and data governance.
For Microsoft, increasing the use of its own MAI models could help lower operating costs while giving the company greater control over the AI capabilities embedded across its software ecosystem. As enterprise AI adoption accelerates, balancing performance with profitability is likely to become an increasingly important priority for the industry's biggest players.
The reported shift does not signal an end to Microsoft's partnerships with either AI company. Instead, it appears to be part of a broader strategy to balance performance with operating costs by assigning selected tasks to its own models while continuing to use third-party systems where appropriate.
The development comes at a time when major technology companies are reassessing the financial demands of generative AI, even as investment in the technology continues at pace.
Microsoft expands its own AI portfolio
Microsoft has spent the past year building a growing portfolio of proprietary AI models under its MAI branding. At its annual Build developer conference last month, the company introduced seven new MAI models, including an AI coding assistant and a text-to-image generation model, signalling its ambition to compete more directly in the foundation model space.
Bloomberg reported that some user prompts in Microsoft 365 applications are now being handled by these in-house models rather than exclusively by OpenAI or Anthropic. Previously, Microsoft had highlighted that many AI-powered features in its productivity suite were driven by models from those external partners.
When asked about the reported changes, Microsoft told TechCrunch that it had nothing further to add.
Developing proprietary models gives Microsoft greater flexibility over how AI services are deployed and may also help the company reduce the costs associated with running high-volume enterprise workloads.
Rising AI costs reshape industry strategies
Microsoft's reported approach reflects a wider trend across the technology sector, where companies are increasingly looking for ways to make AI services more financially sustainable.
After a period of aggressive spending on AI infrastructure and model deployment earlier this year, several large technology firms have reportedly started introducing measures aimed at controlling expenses. Companies including Amazon, Uber, Meta and Accenture have all been linked to efforts to optimise AI-related spending.
The economics of generative AI remain one of the industry's biggest challenges. Training advanced models requires enormous computing resources, while serving millions of user requests every day continues to generate significant infrastructure costs.
The pressure to reduce those expenses has also led some companies to explore lower-cost AI alternatives, including models developed in China. While these options may offer financial advantages, organisations continue to weigh them against concerns surrounding security, compliance and data governance.
For Microsoft, increasing the use of its own MAI models could help lower operating costs while giving the company greater control over the AI capabilities embedded across its software ecosystem. As enterprise AI adoption accelerates, balancing performance with profitability is likely to become an increasingly important priority for the industry's biggest players.
