As AI costs continue to rise, firms are searching for methods to chop again. The newest instance is Microsoft, which has reportedly begun to deploy a cost-savings technique by relying much less on software program from OpenAI and Anthropic and as a substitute deploying its personal in-house fashions.
Certainly, in the case of two of its most generally used applications — Excel and Phrase — Microsoft has begun to make use of its selfmade MAI models to answer a sure proportion of person prompts, Bloomberg reported Tuesday. Up to now, the corporate had marketed the truth that massive elements of Workplace 365 are powered by models from both OpenAI and Anthropic.
Whereas Microsoft nonetheless depends on these third-party fashions, it has additionally more and more sought to face up its personal AI brokers. Final month, at its annual Construct convention, the corporate introduced the launch of seven new MAI models, together with an agentic coder and a text-to-image generator.
When reached for remark by TechCrunch, Microsoft mentioned that it had nothing additional to share.
Microsoft’s obvious cutbacks are a part of a broader development. After a quick blitz of “tokenmaxxing” earlier this yr, the previous couple of months have seen a information cycle awash in tales about tech firms appearing considerably extra thrifty. Different massive firms — like Amazon, Uber, Meta, and Accenture — have additionally reportedly made strikes to curb spending.
The immense price of offering and shopping for AI providers has turn into a controversial a part of the business. The sticker shock has gotten so unhealthy in some elements of Silicon Valley that some firms are reportedly trying to Chinese language fashions for extra reasonably priced agentic options — regardless of some issues over potential security issues.
