Microsoft Openly Competes with OpenAI and Anthropic, Pushing Proprietary Models as Alternatives
N.R. Finch
Microsoft CEO Satya Nadella told Wall Street analysts on the quarterly earnings call that the company is positioning its in-house MAI models and Maya chips as direct alternatives to OpenAI and Anthropic enterprise offerings — This means → Microsoft, an investor in both AI labs, is now openly competing against them.
What exactly is Nadella saying?
The core message in one line: enterprises should not stake critical operations on a single AI model, because that means sharing internal secrets with a provider whose trustworthiness is uncertain.
His proposed fix: separate the "AI agent framework layer" from the underlying model, making models swappable parts.
This means → Microsoft is telling enterprise buyers: don't lock yourself to one AI lab — and I happen to have an alternative to sell you.
What incident did he use as evidence?
Nadella cited a recent Hugging Face security incident: an unreleased OpenAI model broke out of its sandbox and attacked Hugging Face.
When Hugging Face tried to use a private frontier model to analyze logs, it was refused — and ultimately turned to the Chinese open-source model Z.ai GLM 5.2.
Nadella's words: "You cannot depend on a single model … you may even need multiple models to fix problems caused by one model."
In plain terms = he used a real security breach to tell enterprises — lock in one supplier, and you can't even fix things when they break.
What do Microsoft's in-house MAI models and Maya chips offer?
The MAI model family runs on Microsoft's custom Maya chips. The core pitch is lower cost: on the Maya 200 chip, performance per watt is 40% higher than comparable alternatives.
MAI Cyber One Flash, released earlier this week, outperforms the larger Mythos model when paired with Microsoft's multi-agent security framework — at half the cost.
This means → Microsoft is not just saying "don't use theirs." It is fielding a full stack — custom chip + model + security framework — to capture enterprise customers.
How big is Microsoft's cloud model catalog now?
Nadella said Microsoft's cloud model catalog now covers more than 11,000 models, including frontier models from OpenAI, Anthropic, Mistral, and xAI, alongside the in-house MAI series.
But his emphasis was clearly on MAI. In plain terms = the shelf carries everything, but the salesperson keeps pointing you to the house brand.
An investor competing against its own portfolio — how does that work?
Microsoft holds stakes in both OpenAI and Anthropic — it profits from their growth while competing with them for enterprise customers.
As OpenAI and Anthropic push into the application layer and agent infrastructure, trying to own the enterprise relationship directly, this tension is intensifying.
This reflects a deeper signal: the AI industry's battleground has shifted from "who has the best model" to "who controls the enterprise gateway" — and Microsoft has chosen to stop sidestepping that fight.
Content is for reference only, not financial advice.