[NYSE: MSFT] | Current Price: ~$409 | Down ~26% from July 2025 highs
How This Piece Came Together
I wrote the notes for this piece riding shotgun on the drive to TJ Maxx for Memorial Day shopping. That is not a random detail. It is relevant to everything that follows.
I have had a growing thesis on Microsoft for a while now, and the company has become one of my favorite names in the market. That is not a secret if you have seen my portfolio. It is my number one position. But I am becoming more bullish by the day, because I believe Microsoft stands to benefit from AI more than any other company in the world right now.
My brother messaged me today and we were going back and forth about recording devices, note-taking apps, and how easy it would be to record phone calls on an iPhone to use for AI transcription. He has been trying to build a system for himself that transcribes his entire work life across every medium. Phone calls. In-person meetings. Computer audio. Bluetooth. He was manually piecing it together using tools like Plaud, Pocket, HiDock, and LittleBird across a fragmented stack of consumer tools. The conversation crystallized something I had been circling for months.
So I took screenshots of the thread, handed them to Claude, had the whole thing transcribed, and used it as the raw material for this piece. Start to finish, published, in less time than it used to take to find notes from a conversation three days ago.
But here is the thing. Even with the workflow compressed, the biggest friction I ran into had nothing to do with the writing. It was every moment I had to cross a device boundary. Screenshotting twenty conversations on my phone. Finding a way to get those images onto a machine where I could upload them. Feeding them into Claude one batch at a time. Hoping the context survived the transfer intact. That part was still manual, still clunky, and still introduced the possibility of losing something in translation.
None of that friction exists in a world where those messages are already inside the ecosystem. If my iMessage thread lived natively in a platform Claude already had context on, this entire piece collapses from a thirty-minute exercise into something that takes no time at all. The idea surfaces, the context is already there, and the only thing left is the thinking.
That gap between "compressed workflow" and "zero-effort workflow" is exactly what Microsoft is positioned to close. And it is exactly what nobody is pricing into the stock.
The Setup: A Stock Down 26% and a Divided Market
Microsoft has shed more than 26% from its July 2025 highs. The stock now trades at roughly 22 times forward earnings, below its 10-year average of 27x. The selloff has been driven by a few overlapping fears: concerns about whether $190 billion in capex will translate into commercial returns, questions about Azure's growth durability as Google Cloud accelerates, and a broader anxiety that AI will disrupt Microsoft's core Office productivity franchise before they can monetize their own AI investments.
That anxiety recently produced one of the more interesting public debates in the institutional investor world.
TCI Fund Management, run by Chris Hohn, cut its Microsoft position from roughly 10% of the portfolio to about 1% between the end of 2025 and March 2026, unwinding approximately $8 billion in exposure. In a letter to investors, Hohn laid out his reasoning plainly. The pace of AI development is giving rise to new productivity tools that could threaten Office's market dominance. Azure faces intensifying competition. And the capital spending cycle introduces real uncertainty about whether returns will materialize on the timeline the market requires.
Bill Ackman took the other side. Pershing Square began accumulating shares in February 2026 after Microsoft dropped roughly 10% following Q2 earnings, and built what Ackman called a core holding with a disclosed stake of $2.1 billion. His view: the market has become overly concerned about Microsoft's competitive positioning in AI, Azure is growing at 39% on a constant-currency basis, M365 Copilot is monetizing at $30 per user per month across the enterprise, and the stock at 22x earnings is a rare opportunity to own one of the world's dominant technology franchises at a compelling price.
Both arguments are coherent. The market is pricing the Hohn view. I think Ackman is closer to right, and I think both of them are missing the most important part of the thesis.
The Value Stack Nobody Is Pricing Correctly
The debate between TCI and Pershing Square is largely a debate about whether Microsoft gets disrupted by AI or benefits from it. That framing misses the structural position Microsoft is quietly building that makes the question almost irrelevant.
LLMs are becoming commodity infrastructure. I cycle between Claude, ChatGPT, Gemini, and others depending on the task. Technical work gets one model. Creative output gets another. The models are increasingly interchangeable for most use cases. My brother and I talked about this directly in that conversation. He is not loyal to any single model. Nobody serious is.
What is not interchangeable is the context those models operate within.
The iPhone carried our conversation. This site is hosting the output. Claude compressed the workflow between them. But the ideas, and the context that made them worth anything, came from a real conversation between two people who understand the space. The LLM is the engine. Context is the fuel. The platforms governing how that context flows will be worth more than any individual model. That gap is going to keep widening.
A Portrait of the Modern Knowledge Worker
My conversation with my brother was not unique. It is a near-perfect portrait of what motivated, technically fluent people are doing right now in the age of AI.
Unfortunately given the complexity of the platforms in which all the data exists and the hurdles needed to capture the data people tend to resort to jerry-rigging a bunch of disparate tools in order to attempt to do what they would like to actually have done.
He wanted something simple: a complete, searchable record of his professional life across every channel he communicates through. Phone calls. In-person conversations. Computer audio. Bluetooth. He knew what he wanted. The tools to do it natively do not exist yet in one place, so he was doing what smart people do when the market has not caught up to the need. He was stitching it together himself. Researching hardware. Comparing subscription models. Evaluating Plaud against Pocket, reading about HiDock, looking into LittleBird, investigating whether iPhone even allows call recording at all. Hours of research and assembly for a workflow that should take five minutes to set up.
This is the defining behavior of the current moment. Across every industry, in every knowledge-work function, people are doing versions of exactly this. They know AI can help them work better. They know the pieces exist somewhere. And they are spending enormous amounts of time and energy figuring out how to connect those pieces themselves because no single platform has made it seamless yet.
That gap is the opportunity. And Microsoft is closer than anyone else to closing it natively.
What Microsoft Actually Owns
Think about what Microsoft already holds on a typical enterprise employee.
Teams messages. Email. Calendar. SharePoint documents. Org-wide permission sets. Call recordings through Teams Phone. Meeting transcripts through Copilot. Confluence, Jira, internal wikis, Slack via integration. Ambient audio capture through agents running as intermediary apps on Bluetooth headsets.
That is the exact stack my brother was trying to manually assemble with consumer hardware and fragmented apps. Microsoft does not need to assemble anything. They already own the pipes.
Most enterprise employees are living inside this ecosystem whether they think about it that way or not. Outlook, Teams, SharePoint, Office 365. Every Word document, Excel sheet, and PowerPoint deck you are permitted to access but will realistically never read. Data that could have informed better decisions, sitting dormant because no one had the time to surface it. The permissions are already mapped. The org structure is already understood. All Microsoft needs to do is connect an agent to it and they have something no external AI provider can replicate without years of enterprise sales cycles and deep integration work.
And here is what makes the Microsoft version categorically different from what my brother was building for himself. He was capturing his own context. Microsoft captures the entire organizational context.
Not just your emails. Everyone's emails you are permitted to see. Not just your meeting transcripts. Every meeting transcript across your reporting chain. Not just your documents. Every document in every SharePoint library your permissions touch. Not just your calls. Every call routed through Teams Phone across the org, indexed and searchable by your agent scoped to what you are allowed to access.
When you connect that into a single agent, every decision inside that org gets exponentially more informed. A question to your agent is not answered with general knowledge. It gets answered with the full documented history of your organization, your team's current priorities, leadership's stated goals, and the relevant communications from yesterday.
The individual knowledge worker using this natively, out of the box, with no setup, no hardware research, no stitched-together consumer apps, is operating in a fundamentally different way than one without it. Multiply that across an entire organization and the productivity delta becomes a structural one that cannot be replicated by other players in the space.
This is not the first time Microsoft has pulled this move. In the early 2000s, the .xls file format did something similar. Excel became the default language of business data not because it was always the best spreadsheet tool available, but because Microsoft made it the standard and then made everything else speak its language. Once your organization's financial models, forecasts, and reporting lived in .xls, you were not choosing Excel anymore. You were already inside it. The format itself was the moat. Switching meant converting years of institutional work, breaking integrations, and retraining everyone. Almost nobody did.
The context layer Microsoft is building now is the same play at an order of magnitude larger scale. Except instead of locking in file formats, they are locking in organizational memory. The longer an organization runs inside Microsoft's ecosystem with agents capturing and indexing its communications, decisions, and workflows, the more irreplaceable that accumulated context becomes. You are not just switching software. You are leaving behind the institutional knowledge of your entire organization.
The picture we are drawing isn't just another chatbot or transcription app but instead an organizational intelligence infrastructure that I believe is essentially not priced into Microsoft's story at all.
Why the Bear Case Gets It Wrong
Hohn's concern is that AI disrupts Office. New productivity tools emerge. Workflows change. Microsoft loses its grip on the enterprise.
The problem with this view is that it assumes the disruption happens outside of Microsoft's control. But Microsoft is not a passive participant in this shift. They are the landlord of the building where the disruption is happening. Every new AI productivity tool that an enterprise employee uses still runs inside an organization whose data, permissions, communications, and workflows are documented inside Microsoft's stack.
The switching costs here are not just high. They are effectively permanent for any organization that has been inside the Microsoft ecosystem for years. That is most of them.
And if LLMs continue to commoditize, which the trajectory strongly suggests, Microsoft's position actually gets stronger. The value shifts away from whoever runs the best model and toward whoever owns the most complete context layer. Microsoft wins that race by default.
Ackman is right that the fears are overblown. But the deeper point is that the context moat makes those fears structurally irrelevant over a long enough horizon.
The Compounding Advantage
Every Teams call recorded adds to the context layer. Every email processed deepens the organizational graph. Every document created inside SharePoint tightens the knowledge base. Every permission set mapped improves the precision of what each agent can surface to each employee.
The data accumulation flywheel has been running for decades. No new entrant can replicate that starting from scratch. The agent layer being built on top of it is not a new product. It is the monetization of a structural position that already exists.
What people are painstakingly attempting to assembling for themselves, Microsoft is about to hand to hundreds of millions of knowledge workers as a default feature of the software they already pay for. The effort goes to zero. The context goes to maximum. That combination has never existed before at this scale.
The Numbers
Microsoft currently trades at approximately 22x forward earnings against a 10-year average closer to 27x. Wall Street expects revenue to rise 17% in fiscal 2026 after 15% growth in fiscal 2025, with net income expected to climb 26% before normalizing. Azure is growing at 39% on a constant-currency basis and Microsoft has guided for modest acceleration in the second half of the year.
At this multiple, the market is pricing in meaningful disruption to the core business. The context moat thesis suggests that disruption risk is significantly lower than the current price implies, and the monetization opportunity from the agent layer sitting on top of that context is not priced in at all.
Risks
Execution is the primary risk. Having the data and the pipes is not the same as deploying an agent layer that employees actually trust and use. Hohn's argument that AI-driven tools could reshape established workflows and spawn competing platforms deserves respect even if the conclusion overstates the threat.
Regulatory exposure around data privacy and AI governance is real, particularly in Europe. Any restrictions on how organizational data can be used to contextualize agents would slow the flywheel.
Capital allocation is worth watching. Microsoft has one of the largest capital programs in the AI industry, with capex running well ahead of prior guidance alongside a significant stake in OpenAI's for-profit entity. If that spending does not translate into returns on a reasonable timeline, the multiple compression could continue.
The Bottom Line
The TCI versus Ackman debate is a useful frame but it is asking the wrong question. The question is not whether AI disrupts Microsoft. The question is who owns the context layer that all of this AI activity runs through.
I believe Microsoft will being to capitalize on this context layer very soon. This will come out of the box as part of the Office 365 suite bundle with no hardware, no additional subscriptions, and no assembly required. And when it does work, it will not just have your context. It will have your entire organization's context, scoped to your permissions, surfaced through an agent that understands your role, your priorities, and your history inside the company.
The individual productivity gains will be meaningful. The organizational gains will be compounding and durable. And the switching costs that come with years of accumulated context inside Microsoft's ecosystem are permanent.
Microsoft is not an AI story in the way most investors are framing it. It is a context monopoly story. The AI wave does not disrupt their position. It makes their position more valuable.
The companies that own the context own the outcome. At 22x forward earnings, down 26% from its highs, you are getting a chance to buy that position at a price that reflects the bear case. The bull case is not priced in at all.
That is the secret sauce. And it is already in place.