Microsoft's AI business depends on one customer more than most investors probably realise. OpenAI accounts for roughly 70% of Microsoft's AI revenue and around 45% of its commercial backlog, according to the company's latest disclosures. That concentration is becoming the latest test for a stock market that has spent the year moving from one AI anxiety to another.
Earlier worries about overinvestment in data centres have subsided for now. Microsoft, Amazon and Alphabet have all reported strong growth in cloud computing revenue, suggesting the heavy spending is generating returns. The new concern is where that growth is coming from.
For Microsoft, the answer is mostly OpenAI. Elsewhere, Anthropic is thought to be doing much of the heavy lifting. Across the sector, growth is being carried by two companies and not much else.
Why customer concentration matters
Both OpenAI and Anthropic face structural problems that could limit their ability to keep signing giant compute agreements. The biggest issue is cost. Most AI use does not need the most advanced model; something good enough at the lowest price will do, and Western labs operate at a major cost disadvantage, partly due to higher energy prices.
Where frontier capability does matter, staying at the front is expensive and difficult. The leading model appears to change almost every week. If OpenAI and Anthropic cannot maintain a sustainable advantage, their demand for cloud capacity could shrink, and that would hit the revenue growth justifying AI investment across the sector.
Microsoft's hedge
Microsoft appears acutely aware of the risk. It is positioning itself as a platform that routes queries across different models rather than committing to one. Satya Nadella appears to back that approach for the longer term.
The immediate danger is that the strategy pressures the one customer driving Azure's growth. Microsoft is also a major shareholder in OpenAI, which deepens its exposure. If the platform pivot succeeds, though, Microsoft could hold a strong position regardless of which model wins. Being a model-neutral interface between users and large language models is not a strategy every company in the AI ecosystem can pull off.
What could go wrong
AI has been the major force pushing stock prices higher, and so far nearly every company in the supply chain has benefited. If that support unwinds, a broader market decline is possible. There are signs the ground is shifting. Some companies are enjoying a short-term AI boost that is likely to fall away, and that could reveal itself sooner than investors expect.
Why this matters for finance professionals
The number to watch is whether AI revenue broadens beyond OpenAI and Anthropic. Customer concentration at this scale can look sustainable until it is not. A slowdown in OpenAI's spending, a change in model leadership, or a price war on inference would show up quickly in Azure's reported growth. Companies that operate as neutral infrastructure rather than depending on one model provider's fortunes are better placed to hold up if AI-led growth narrows. That distinction matters when assessing which stocks can weather the next phase of the AI cycle.
Your membership also unlocks:
