The MoatWide moat

Microsoft may be the finest toll-taking business in all of technology, and the key to understanding its moat is to notice where it sits: not in front of the customer, where fashions change and loyalty is fickle, but underneath the customer, in the plumbing of the working world. The operating system that boots the office computer, the email that carries the company's correspondence, the documents in which its work is actually done, the identity system that decides who is allowed to log in — Microsoft supplies all of it, and it charges rent for all of it. In recent years it has converted that rent from a one-time license into a recurring subscription, which is better still, because a subscription is an annuity that renews itself while a license is a sale you must win again.

The genius of the position is that once a large enterprise has built its daily processes on Microsoft's rails, ripping them out is a project no sensible chief information officer volunteers to lead. The switching cost is not a fee written into a contract; it is the accumulated risk, disruption, and retraining that a migration would inflict on thousands of employees who know no other way to work. So the customer does not really shop at renewal time. The customer renews a dependency, and Microsoft raises the price a little, and cross-sells one more product, with a confidence very few businesses on earth ever enjoy.

On top of that dependable base the company built a second great franchise in Azure¹ — a genuine, global-scale cloud with the capital, the engineering, and the enterprise relationships to keep pace with the largest players in the world. And because those enterprise customers already trust Microsoft with their software and their identity, moving their computing to Microsoft's cloud is a short and natural step rather than a leap of faith. The company bundles the whole thing with a salesman's cunning, so a customer already paying for Office and Windows finds the cloud, and now the artificial-intelligence tools layered on top of it, an easy and logical 'yes.' The newest toll on that same road is artificial intelligence: Microsoft has moved to layer AI assistants across the very tools its customers already open every morning — the documents, the email, the chat, the developer's editor — and to charge a fresh subscription for each. Because the customer is already inside the house, the cost of introducing that next product is close to nothing, and every enterprise already paying Microsoft is a ready prospect for the next thing it decides to sell.

It is easy to forget, amid the software and the cloud, just how broad the empire has become, and the breadth is itself a kind of moat because it spreads the company across so many independent sources of strength. Microsoft owns LinkedIn, which is the professional identity of much of the white-collar world and a network effect in its own right; it owns a large gaming franchise in Xbox and its studios; it owns the search and advertising business behind Bing. No one of these would define the company, but together they mean a stumble in any single market is cushioned by the others, and each provides one more surface on which to sell the next thing to a customer already inside the tent.

Add to all this a developer platform — the languages, the tools, GitHub, and now the AI coding assistants² — that a generation of software builders learns on, and you have switching costs layered upon switching costs upon a network effect. The risks are real and worth naming: regulators have long eyed the company's dominance, open-source software nibbles at the edges of its franchises, and the shift to artificial intelligence could, in principle, rearrange the competitive landscape in ways no one can fully foresee. But it is telling that Microsoft has so far been among the chief beneficiaries of that shift rather than its victim. This is about as embedded, as diversified, and as sticky as a software business ever gets, and embeddedness of that degree is a moat you can very nearly touch. If you want the one number that tests it, watch the commercial backlog — the $678 billion of contracted, not-yet-delivered spend³, which grew 84% in a year against revenue's 18%. Enterprises do not sign multi-year commitments of that size with a vendor they are edging away from; the year that backlog stops outgrowing revenue is the year the grip has begun to loosen.