Microsoft (MSFT) Is Essentially A Utility
Everyone is a momentum trader these days. The active management houses are playing follow the AI leader. How can one be expected to outperform over the long-term if all he/she does is chase performance?
Take the example of Microsoft (MSFT). If Satya Nadella was forced to disclose to investors the contribution to revenue growth from price increases vs. unit growth, I suspect investors would be greatly disappointed.
For example, if you’re on MSFT’s 365 Subscription, you have had the pleasure of seeing your plan subscription price increase 30% in 2025. One could retort: “Yes, but MSFT has not lifted 365 prices for a few years..” to which I say: “Sure, but where’s my increase in related features and functionality?”
It’s poor 365 customer relationship management to increase 365 customer prices simply because your input costs on the Azure side are moving higher as a result OpenAI uptake (although GPT 5 is inferior to Claude 4.1 Opus), and higher NVIDIA (NVDA) chip costs. Make Azure AI customers eat the full cost rather than have 365 subsidize a portion of Azure’s OpEx.
MSFT’s AI apps effort (Copilot Pro), has been a miserable failure as we have written about on numerous occasions last year (pulling Copilot Pro numbers after only 2 quarters was a weak move and poorly-coordinated financial disclosure).
At the end of the day Microsoft is an AI back office. Sure, MSFT has a nascent chip design effort and its own small language model research effort, but the cutting edge AI work is being performed at Nvidia (NVDA), Groq, Anthropic, Google DeepMind (GOOGL), OpenAI and Grok/xAI (One could include Meta (META), but Zuckerberg shifting his focus every few years - crypto, metaverse/AR, now AI - isn’t going to attract an army of the best and brightest, even if he overpaid for Ilya Sutskever’s CEO).
Back to Microsoft. What does one pay for a staid Productivity Software developer that isn’t doing much of anything on the Software Development front while serving as the back office for OpenAI?
Remember when Sumner Redstone said that “Content is King”? The chip designers and model builders are the content, while Microsoft provides the IP-generic distribution. Microsoft’s business model (IP-lite), by definition does not warrant a hefty EV/EBITDA or P/E multiple. A 38x P/E is rich for MSFT given its 15% EPS and 15% Operating Cash Flow growth, not to mention its single-digit percentage Free Cash Flow growth, which is essentially being spent on content delivery - i.e. a utility-like business.
We will publish a brief piece soon about how the LLM space will shake out in 2026, given LLM performance is no longer growing by 10x despite a 10x increase in training costs. The $100 billion model that Dario at Anthropic spoke of isn’t going to happen in 2026. I suspect that some OpenAI and Anthropic investors may push for an exit soon while a story can still be spun that would pull the wool over institutional investors’ eyes.