streugut.ch
Archive

AI competition is moving away from frontier models alone toward control of the stack that turns models into governed execution. That shift is being carried by agent environments tied to enterprise identity, data, workflow, security, pricing, deployment services, and everyday work surfaces, and by model access that is increasingly routed through managed clouds, owned product surfaces, and access controls. The infrastructure race underneath is widening into financing, datacenter capacity, packaging, custom silicon, and power. What is particular now is that this execution layer is becoming a make-or-buy boundary as well, with platforms widening customer model choice while quietly pulling more workload inside their own commercial and operating systems.

The move

Microsoft begins routing some Word and Excel prompts to its own MAI models

The clearest new move came on July 7, when TechCrunch, citing Bloomberg, reported that Microsoft has begun routing a portion of Word and Excel user prompts to its in-house MAI models instead of relying solely on OpenAI and Anthropic systems. That changes which models sit behind part of Microsoft’s core productivity software.

That came just after Microsoft said on July 6 that it is eliminating around 4,800 roles, redeploying over 4,000 employees into new roles, and restructuring Xbox and commercial sales. The same day, Microsoft said Moonshot AI’s Kimi K2.7 Code model is available in public preview on Microsoft Foundry.

Among the main competitors, OpenAI and Databricks said on July 6 that they are working together to help enterprises move from prototype to production AI agents. Anthropic said on July 7 that it has begun rolling out its Claude Cowork tool in beta on the web and mobile for Max subscribers. Google said on July 7 that third parties can package and sell AI agents through Gemini Enterprise and Google Cloud Marketplace.

Sources