Microsoft

Microsoft started as an operating systems company - the software that makes the computer itself work before you even open a program. Windows is that operating system, installed on most office and home machines around the world. Office is the package with Word, Excel and Powerpoint - the words, tables and presentations of whole generations of employees pass through it. The two products carried the company for decades, even as the world turned to phones and the cloud.

Then came Azure - a cloud platform, meaning other people's servers that you rent instead of buying your own machines. Azure is exactly why Microsoft sits where the future of artificial intelligence is being decided now: OpenAI's models are trained and run on it. It also bought GitHub - the place where programmers keep and share code, practically the archive of modern software - and LinkedIn, the professional social network where people show off their career instead of vacation photos.

Since 2019 Microsoft has been pouring money into OpenAI - committing 13 billion dollars in investment for a stake in the company and, more importantly, for OpenAI's right to use Azure exclusively for its computing. After OpenAI's restructuring in 2025 into a more conventional commercial form, Microsoft's stake is estimated at about 27 percent. Not a majority, but enough to be the largest outside owner of a company valued in the hundreds of billions.

Here's the absurd knot. Microsoft sells OpenAI the computing power it can't exist without - meaning it's its supplier. At the same time it's a co-owner, meaning it profits if OpenAI succeeds. And third, it builds OpenAI's technology into its own products - Copilot, the assistant that writes text and code inside Office and Windows - while in parallel developing its own models so it doesn't depend on anyone. Supplier, owner and competitor of the same thing, all at once.

Under Satya Nadella, CEO since 2014, the old desktop software company is redressing itself as a cloud and AI company. The move with OpenAI wasn't a random bet - it's insurance. If generative artificial intelligence becomes the new main interface to the computer, Microsoft wants to be both the treasury behind it and the shelf the finished product sits on.

The insurance doesn't rest on OpenAI alone. On October 9, 2026 Microsoft released Microsoft-Decision-1, a decision model: given a set of options, it returns a probability for each one - yes or no, a choice, a rating, including rubric-based grading of AI responses and agent actions. It's available as a public preview in Microsoft Foundry and through OpenRouter, at $0.042 per million input tokens, with output free. The foundation belongs to neither Microsoft nor OpenAI: the model is a further-trained Qwen3.5-9B, an open model from Alibaba Cloud's Qwen family. Moving it onto Microsoft's own models (MAI) and onto OpenAI's is promised "soon", with no date. The accuracy and speed numbers are Microsoft's, from its own tests, among them 35 times faster than GPT-6 Sol; there's no independent evaluation yet.

Microsoft didn't bet on OpenAI. It bought the ground it stands on, and the chair it's watched from.

I'll believe it when I see it

First I watch not the actor, but the set behind him - who built the stage, where the power for the lights comes from. Here the set is Azure, the actor is OpenAI. The audience shouts the actor's name, but the check for renting the hall, the lighting and the electricity goes into the same till. It's not a conspiracy. It's an architecture of dependence, made explicit and legal.

The old empire didn't buy a ticket to the future. It bought the very hall where it will be staged. And when the play becomes too successful to control from outside, it simply steps backstage and becomes part of the production. That's Microsoft.

The visual is generated code-art, no third-party images.