Microsoft Reportedly Reduces OpenAI and Anthropic Reliance as AI Costs Rise

Microsoft is reportedly increasingly using its own AI models in some products, amid rising concerns about the financial cost of relying on powerful external AI systems. For some of the requests that used to be handled by technology from companies like OpenAI and Anthropic, the company has started to use internal models. The change is not a complete separation from either company. Instead, Microsoft seems to be creating a flexible system where you can pick different models based on performance, complexity, speed and operating cost.

Also Read

HMD Skyline 2 Leak Reveals Snapdragon 7s Gen 4 Chip Ahead of the Official Launch

Microsoft Grows Its Own MAI Model Family

A lot of this strategy is centred on Microsoft’s growing number of internally developed MAI models. The company released seven models that cover reasoning, coding, image generation, voice and transcription. MAI-Thinking-1 is trained for complex reasoning and long-context tasks, while other specialised models are tuned for other types of work. Developing its own technology to support a broader mix of AI workloads will help Microsoft avoid the need to send every request to expensive outside systems.

Also Read

Sony RX10 V Brings Major Camera Upgrades With a Premium $2,299 Price Tag

Anthropic and OpenAI Continue to Be Critical

Microsoft isn’t ditching its big AI partners. Copilot has increasingly been taking a multi-model approach combining technology from OpenAI and Anthropic. Features like Critique, where one model generates content and another assesses its quality, and Council, where users can compare the answers of different systems, are examples of this. This is a clear signal that Microsoft is heading towards model choice rather than total reliance on one AI provider.

Also Read

SiriusXM Outage Disrupts Service for Thousands of Users as Downdetector Reports Widespread Problems

Why Higher AI Costs Are Changing Product Choices

As companies expand the use of assistants and autonomous agents, the economics of generative AI are becoming more important. More complex AI workloads can consume lots of tokens, especially when the software needs to reason for a long time or do a sequence of tasks one after the other. At Microsoft’s scale of consumer and enterprise products, the cost of processing each request can vary dramatically. This provides a strong motivation for using efficient models for routine jobs, and saving expensive systems for difficult tasks.

Also Read

Solos AirGo A6 Smart Glasses Launch With New Features Designed for Everyday Use

Microsoft wants more control over the future of AI.

Microsoft’s in-house model development also provides the company with more strategic independence. The competitive AI technology enables Microsoft to make product decisions without having to consider the pricing, availability or development plans of another company. Microsoft AI leadership has been increasing its internal research and model development, while continuing access to external AI platforms. The strategy may enable the company to react faster as the capabilities of the models and the business relationships change.

Sources

• • TechCrunch – Used in opening paragraph about Microsoft increasing its use of proprietary models to cut costs of running AI and lessen dependence on external providers.
  • Microsoft AI – For the MAI model paragraph, the seven-model family and focus on reasoning, coding, image, voice and transcription tasks.
  • Reuters Copilot – will continue to use the OpenAI and Anthropic models, but now it will do so through the multi-model features.
  • Fortune – Talking about token consumption, inference costs and the broader AI cost problem.
  • The Verge – Used to develop Microsoft’s internal model and its growing ambition to build more powerful independent AI capabilities.