$Microsoft(MSFT.US)'s current pricing is based on pessimistic expectations for OpenAI. OpenAI, which was deified last year, has been surpassed by Anthropic this year, seemingly disproving the commercial closed-loop of the to-C large language model: last year, the capital market could tolerate it burning money and raising funds regardless of cost, but against a not-so-optimistic macro backdrop, the requirements have become stricter. The dilemma of the to-C large language model is that subscription revenue alone is insufficient to offset data center costs; while adding ads would make the user experience extremely poor. In contrast, Anthropic, which focuses on the to-B side, has come from behind. OpenAI has had to close down money-burning business lines like Sora. At the same time, the binding between Microsoft and OpenAI is gradually weakening.

Microsoft, as one of the largest B-side enterprises, will not perish because AI is shifting to the B-side. The switching cost for enterprises using Microsoft products remains its deepest moat. Azure, as a native AI Cloud, is not only bound to OpenAI; the future AI business model is likely to evolve into the sale of Tokens. Microsoft, as the underlying infrastructure of the AI era, will not be replaced. Large models are easily distilled, but data centers and computing power are not.

Note that among the major US tech giants, Microsoft's gross margin (68%) is second only to $NVIDIA(NVDA.US) (75%). And regarding 2026 capital expenditure on AI, among the three major cloud vendors, Microsoft is the lowest.