September 1, 2026: GitHub changes billing as Microsoft’s AI buildout stays the bigger question

Today is September 1, 2026, and for Microsoft the most visible new change is not a quarterly number, but a GitHub policy shift that starts to tighten how Copilot is sold and which models it can use.

GitHub is reopening Copilot Business and Copilot Enterprise sign-ups today. It will require payment before seat access, and it will start upfront seat charges for existing credit-card and PayPal customers on October 1. At the same time, several Copilot model options are being deprecated today, including Gemini 3.1 Pro, Claude Opus 4.5 and 4.6, Claude Sonnet 4.5 and 4.6, and Raptor Mini. Raptor Mini is shifting to MAI-Code-1-Flash.

That matters for two reasons. First, it is a direct monetization and access-policy change on a core developer product. Second, it shows Microsoft pulling more of the control points closer to its own stack. When a platform narrows the model menu, it can simplify support, shape quality, and potentially manage inference costs more predictably. The tradeoff is that users and enterprise admins may watch closely for any effect on flexibility or performance.

There is also a broader Microsoft 365 Copilot update from August 25. The release notes say Microsoft added Python in Excel via Edit with Copilot and increased Copilot Studio Agent Builder file uploads to 512 MB. That is a concrete workflow improvement, not a headline about demand by itself, but it does matter because it widens the range of tasks users can hand to Copilot inside familiar Microsoft tools.

On Windows, Microsoft’s latest on Arm note says the app catalog now has more than 7,000 verified apps and games. It also says fall 2026 devices from Surface, ASUS, Dell, HP, Lenovo, and MSI will join the platform with NVIDIA RTX Spark. The plain-English takeaway is that app availability is still the key commercial limiter for Windows on Arm. More verified apps reduces one of the main reasons buyers hesitate.

There is a similar distribution story in education. Microsoft and DepEd said they will give Microsoft Copilot access to 1 million Filipino public school teachers, and that the skilling program has already reached more than 175,000 educators. That is a large deployment signal, but it is still a company claim about access and training, not a revenue disclosure. It is best read as distribution and familiarity building, not proof of monetization.

Microsoft also announced a collaboration with HUMAIN to bring ALLAM models into Microsoft Foundry and the Microsoft 365 Copilot ecosystem, with forward-deployed engineers for implementation support. That extends Microsoft’s AI stack deeper into Arabic-language and regulated enterprise workloads. Again, the important point is not the announcement alone, but where it fits: Microsoft is trying to make its platform useful in more local, specialized, and operational settings.

For Xbox, Insiders can start testing the new disc-to-digital entitlement feature as of August 31. That starts with most Xbox One and Xbox Series X disc games and thousands of titles. It is a gradual move toward digital ownership and engagement. It does not tell us gaming economics by itself, but it does show how distribution is being reshaped at the product layer.

The most important thing to remember is what we do not yet have. Microsoft’s latest filed financial anchor is still the July 29 FY26 Q4 8-K and earnings release. In this pass, there is no newer SEC filing updating Azure growth, revenue, commercial backlog, or any new guidance. So the core scorecard remains unchanged: Azure capacity versus demand, Copilot monetization, and datacenter buildout are still the key unanswered questions.

Here is the finance mechanic to keep in mind. Microsoft can rent or finance some of the hardware and buildings it needs instead of buying everything outright. That can bring capacity online sooner, especially for AI infrastructure, but the cash and accounting costs still show up over time. So when you hear about faster capacity expansion, it is worth asking whether the company is also leaning more on datacenter finance leases, and how that shows up later in the numbers.

So the near-term read is simple. Microsoft’s product teams are still pushing harder on distribution and monetization, especially in GitHub and Microsoft 365. But the investor-relevant question is still whether those moves translate into measurable changes in cloud growth, Copilot usage, and capital intensity.

What to watch next is concrete. First, whether Microsoft or GitHub publishes any follow-up on Copilot seat adoption after today’s billing change. Second, whether any later filing or earnings material updates Azure growth, capex, or datacenter lease usage. Third, whether Microsoft gives any usage or revenue signal tied to Microsoft 365 Copilot or GitHub Copilot after the billing transition begins.