Despite being an investor in OpenAiand one of its main collaborators, everything indicates that Microsoft also wants to become one of its competitors. To do this, he is developing various AI models, in addition to being valuing alternatives to OpenAi’s products and tools to incorporate their products, among which is its Copilot bot.
Microsoft has already developed, according to The Information, its own models of AI reasoning, with results comparable to those that have reached O1 and O3 Mini of OpenAi. And what seemed slight tensions between both companies have accentuated after Openai apparently wanted to answer Microsoft’s questions about the technical details of the O1 model.
In addition, Microsoft has developed a family of models, which has been called MAI, who can compete with those of Openai. According to Bloomberg, he is seriously valuing to offer them through his API before he finishes this year. Also, apparently, those of Redmond are testing alternative models to those of Openai, from companies such as Meta, Anthropic and even Deepseek to incorporate them to Copilot as substitutes for OpenAi technology.
However, the advances in this area have not been very agile, and in the company there have been several obstacles to make the change. It is unknown what kind of barriers can be, which could range from technical difficulties to internal resistance from Microsoft departments to change.
After having invested around 14,000 million dollars to date in Openai, he is trying to diversify in various ways. To this end, one of the founders of Deepmind, Mustafa Suleyman, which he has put in charge of his AI related areas, which has grouped into the Microsoft AI division, has signed for this purpose. Your mission? Reduce the dependence that the company has of OpenAI
Apparently, Microsoft wants more control over costs and also on its own models. The costs of executing Openai models, among which is GPT-4, the basis of Microsoft’s co -ilot assistant, are not precisely economical to execute. Keeping them in execution raises the costs remarkably, and those of Redmond want to reduce this invoice with cheaper alternatives.
As is obvious, Microsoft wants to have its own reasoning models, which apparently began to plan after Openai’s internal crisis, in which Sam Altman was expelled for a few days of his position and the company. This incident demonstrated Microsoft’s vulnerability if its partner has serious internal problems.
But achieving it is more complicated than it seems. To start, Microsoft and OpenAi have a very strong relationship. The second technology is integrated into Microsoft products, from Co -cilot to Bing searches or Microsoft 365 productivity tools.
A change in the agreement that the two companies adopted last January allows Openai to work with other cloud suppliers, but Microsoft still maintains exclusive rights over OpenAi models for their products until 2030, a fairly long period.
The agreement, until now, has also been quite lucrative, since Microsoft stays with a part of OpenAi’s income, but not without tensions. In addition, the losses that Openai has had 2024 have made Microsoft bet more firmly for having their own agenda and AI systems.
And The Redmond finally get rid ofat least in part, of OpenAithey will have many advantages, such as possibility of developing faster and more economic services. Also the possibility of marking its territory in a very competitive sector. But among the contractual obligations they have with OpenAI, the technical dependencies and their continuous innovation, the advances to become independent can take a lot of effort and time to Microsoft.
