There is a remarkable change in Microsoft’s approach to artificial intelligence. According to information provided by Bloomberg, the company started to increase the share of its own developed models in the artificial intelligence infrastructure used in Office applications. Especially in Word and Excel, the MAI models developed by the company now respond to users’ requests from a certain section. To date, Microsoft has clearly emphasized that many artificial intelligence features in the Office 365 ecosystem are supported by large language models developed by OpenAI and Anthropic. Although the new application does not completely end Microsoft’s dependence on third-party models, it shows that it is gradually changing the balance in favor of its own technologies.
Microsoft continues its long-standing strategic partnership with OpenAI. Despite this, the company has recently been devoting more resources to developing its own artificial intelligence models and agents. This approach is not only intended to keep costs under control; It also makes it possible to achieve greater flexibility in product development processes and to manage the artificial intelligence infrastructure more effectively within the company. In addition, the company aims to make its commercial products more independent by developing special models for different usage scenarios.
Microsoft expands use of its MAI models
Microsoft introduced seven new artificial intelligence models of the MAI family at the annual Build developer conference held in May. These models include an agent-based coding model that can help software development processes and an artificial intelligence model that can produce images from text. Although the company does not share details about which products and on what scale these models will be used, according to Bloomberg, Word and Excel have become one of the first areas of use. Microsoft did not make an additional statement in response to TechCrunch’s request for comment on the issue.
This step by Microsoft coincides with a broader trend in the technology sector that has become evident in recent months. While the “tokenmaxxing” approach, which refers to allocating more processing power and more tokens to artificial intelligence models, was widely discussed in the first months of the year, it seems that companies are monitoring expenditures more closely recently. As the operating costs of artificial intelligence services increase, technology companies began to turn to lower-cost alternatives. For this reason, not only model performance but also the balance of cost and efficiency becomes more decisive in decision processes.
On the other hand, similar cost control steps are not limited to Microsoft alone. News has been published that large companies such as Amazon, Uber, Meta and Accenture have planned their artificial intelligence investments more carefully and made various arrangements to reduce operational expenses. Developing and operating artificial intelligence services on a large scale creates a significant cost due to high hardware investments, energy consumption and cloud infrastructure expenses. This situation leads companies to develop their own models or use existing solutions more efficiently.
In addition, interest in lower-cost artificial intelligence models is increasing in the industry. Some artificial intelligence models, especially China-based, have begun to attract the attention of global companies on the grounds that they offer more cost-effective agent-based solutions. However, choosing these models brings with it various questions in terms of security, data privacy and regulatory requirements. For this reason, many organizations make decisions by considering not only the cost but also data security and corporate compliance criteria.
The fact that Microsoft includes more of its own MAI models in its Office applications reveals that the company attaches more importance to the balance of cost and control in its artificial intelligence strategy. While the company continues its collaborations with OpenAI and Anthropic, it aims to reduce external dependency by using its own technologies more widely in its products. Whether a similar approach will be reflected in Microsoft’s other services in the coming period will become clear with new announcements to be made by the company.