Microsoft sold the same assets three times: as licences, as subscriptions, as cloud

Microsoft’s portfolio can be told as a sequence of reuses. Windows gave the company a platform on personal computers, and Office became its flagship product, sold as a licence that a customer bought once and upgraded every few years. In June 2011 Office 365 put the same applications on a monthly or yearly subscription. Azure, Microsoft’s cloud, became commercially available in 2010 and turned the company’s experience of running servers into capacity sold by consumption. Xbox Game Pass applied the subscription logic to games from 2017.

The latest layer is AI. Microsoft 365 Copilot became generally available to large business customers on 1 November 2023, at 30 dollars per user per month. On 29 July 2026, closing its fiscal year, Microsoft reported that Azure revenue had passed 100 billion dollars for the first time in the year and that Microsoft 365 Copilot had more than 30 million paid seats. The record also has a different column. Zune, launched in 2006, was discontinued in 2012. Windows Phone never gained a large share, and in 2015, a year after buying Nokia’s phone business, Microsoft took a charge of about 7.6 billion dollars related to it.

A climb that starts from a platform

In our pattern library this shape is the Climb: each new layer reuses the customers and assets of the layer below, so the moves rhyme by construction. Microsoft runs the platform-first version. The platform came first (Windows), the product sat on it (Office), the same product became a recurring service (Microsoft 365), and the company then reached down to own the ground underneath, the computing capacity that its own services and its customers’ software run on (Azure).

Each step changed the price model more than the product. A customer who once paid for Office every few years pays every month, and a company that once bought servers pays for what it uses. The same assets, sold as licence, then subscription, then consumption, earn more often and leave the customer less reason to look elsewhere at each renewal.

Copilot as a layer across everything

Copilot follows the same rule. It is sold as an addition to subscriptions customers already hold, runs on Azure’s capacity, and lives inside the applications people already open every day. That makes it the most rhymed kind of launch: a new service with an installed base waiting for it. The 30 million paid seats reported in July 2026 come from that base, and so will any growth. The risk is the other side of the same structure. An AI layer has to prove its value one task at a time inside tools customers already pay for, so the pricing question is whether the extra charge shows up as hours saved that a manager can see.

The orphans prove the rule

The launches that failed were the ones that could not reuse anything Microsoft already sold. Zune entered a music player market Apple’s iPod already led, with no Windows advantage to carry it. Windows Phone arrived after the iPhone and Android had set the rules, and the developers who built for Windows had less reason to follow than the size of Windows suggested. In the books these are Orphans: launches with no part of the existing portfolio ready to catch them. The Surface line, by contrast, survives as a showcase for Windows, a product that echoes the platform it runs.

Read together, the climb and the orphans give the same advice. A new offer has the best chance when the customer can reach it from something they already pay for, and the worst chance when the company’s strengths stop at the door of the new market.

If your next launch needs to stand on what you already sell, map which customers and assets it can reuse before you price it. Scoring that is part of what we do.