Everyone's Terrified Microsoft Will Keep Spending. I'm Terrified They'll Stop.

Microsoft just burned $37.5 billion in 90 days. The stock has lost a third of its value. And the most important number on the balance sheet is the one nobody's talking about.

Thirty-seven point five billion dollars.

That's how much Microsoft spent on capital expenditures last quarter. One quarter. Ninety days. Enough to buy every cattle ranch in the state of Nebraska, twice, and still have change left over for a reasonably nice steak dinner.

Wall Street saw that number on January 28th and did what Wall Street does. It panicked. The stock dropped 10% the next morning. It has kept sliding since. As I write this, Microsoft trades at $356.77, down 34% from its October high of $539. The market has erased roughly $1.3 trillion in value in five months.

The headlines write themselves: "Microsoft's AI Gamble." "The $150 Billion Bet That Could Backfire." "Is Satya Nadella Burning Cash?"

If you've been investing for any real length of time, you've watched companies incinerate capital on fiber optic cable in 2000, on subprime derivatives in 2007, and on cryptocurrency exchanges in 2022. Another round of "this time is different" triggers a very rational immune response. I feel it too.

But I've spent the last week pulling apart Microsoft's actual filings. Not the analyst summaries. Not the headlines. The 10-Q, the earnings transcripts, the segment disclosures, the cash flow statements. And what I found is a business so fundamentally healthy that the current panic looks less like prudent skepticism and more like a misreading of the financial statements.

The crowd is staring at the CapEx line. They should be reading the balance sheet.

The Cash Machine Nobody Mentions

Microsoft just completed a fiscal year in which it earned $101.8 billion in net income. That is not a typo. One hundred and one point eight billion dollars in profit. No software company in the history of capitalism has ever done that.

Revenue hit $281.7 billion, up 15% from the year before. Operating income reached $128.5 billion, a 17% increase. The operating margin expanded to 45.6%. Think about what that means. For every dollar that flows through Microsoft's front door, forty-five cents and change falls straight to the operating line before the tax collector arrives.

Now, free cash flow. This is the number I care about most because it represents the actual cash a business generates after it has paid for everything it needs to keep running. Microsoft's came in at $71.6 billion. That is not revenue. That is not operating income. That is cash left over after paying every employee, every supplier, every tax bill, and every capital expenditure. Seventy-one billion in cash that the company can do whatever it wants with.

Here's the part that genuinely surprised me, because it contradicts the narrative you're hearing on financial television. Microsoft holds $89.4 billion in cash and short-term investments against roughly $40 billion in total debt. Read that again. The company could write a check tomorrow to pay off every cent it owes and still have $49 billion left in the account.

Microsoft is not leveraged. It is not stretched. It has a net cash position of $49 billion. The company paid about $24 billion in dividends last fiscal year. It could cover nearly four years of dividend payments from its cash reserves alone, without earning another dollar. Most companies can't cover four months.

There's an old-fashioned test called the Piotroski F-Score. Nine points. It checks whether profits are real, cash flow exceeds earnings, and leverage is declining. Simple stuff that has held up for decades. Microsoft scores a 7 out of 9. The two missed points? Gross margins ticked down slightly because of AI infrastructure costs, and asset turnover declined because the company is building data centers faster than revenue can fill them. Both are consequences of investing, not deteriorating.

A business that generates $71.6 billion in free cash flow, maintains a 45.6% operating margin, earns roughly 28 cents of profit on every dollar of invested capital, and holds a net cash position of $49 billion is not a business in crisis.

It is a business that Mr. Market has decided to put on clearance because the CapEx line got too big for a spreadsheet.

The Spending Is the Signal. Not the Problem.

Here's the part the frightened money doesn't understand.

Microsoft isn't spending $37.5 billion a quarter because Satya Nadella lost his mind. It's spending $37.5 billion a quarter because customer demand for Azure cloud computing is growing at 39% per year and the company literally cannot build servers fast enough.

Sit with that for a moment. Azure, Microsoft's cloud platform, grew 39% in the most recent quarter. Amazon's AWS, the market leader, grew 24%. Google Cloud managed 48% but from a considerably smaller revenue base, so the absolute dollars tell a different story. In the cloud infrastructure market, which now exceeds $400 billion annually, Amazon holds 28% share, Microsoft holds 21%, and Google holds 14%.

Those three companies control two-thirds of the market. The barriers to entry are almost comically high: tens of billions in capital, a global network of data centers, deep relationships with enterprise procurement departments, and armies of engineers who don't exactly grow on trees. No new competitor has gained meaningful share in a decade. This is not a market. It is an oligopoly. And Microsoft is the fastest-growing member.

When a company with these economics tells you it is spending aggressively to meet demand that exceeds supply, the correct emotional response is not fear. It is gratitude. You are watching a toll bridge operator widen the highway because traffic has doubled.

The moment Microsoft says "we've decided to slow down our infrastructure spending because demand is softening"? That is the day to be frightened. That would mean the AI workload growth story has cracked. That would mean the enterprise customers writing hundred-million-dollar Azure commitments have pulled back. That would mean the $625 billion in commercial backlog is evaporating.

None of those things are happening. The opposite is happening.

Microsoft brought one gigawatt of data center capacity online in a single quarter. And it still wasn't enough. The company's CFO told analysts that capital expenditures will decrease sequentially in the next quarter, and before you get excited, it's not because demand is weakening. It's because of "normal variability in the timing of delivery of finance leases." In plain English: they are constrained by how fast they can physically plug buildings into the power grid.

I've been studying capital allocation decisions for most of my adult life, and I cannot recall a single instance of a company spending $37.5 billion in ninety days and having its Chief Financial Officer essentially tell analysts: "We wish we could have spent more."

Three Moats, Stacked

Buffett has always said he wants to own businesses surrounded by a moat filled with piranhas. Microsoft's moat is unusual because it's really three moats stacked on top of each other, and they feed into one another in ways that I think the market underappreciates.

Start with switching costs. More than a billion people use Microsoft 365. Every Fortune 500 company runs on Active Directory, Outlook, Excel, and Teams. I've talked to CIOs who have tried to migrate off this stack. The cost isn't measured in licensing fees. It's measured in years. Years of lost productivity, retraining, integration headaches, and the career risk that falls on whoever signed off on the project. Nobody does it unless they're absolutely forced to. This is why the Productivity and Business Processes segment grew 13% to $120.8 billion, even in a year when plenty of companies were tightening every other budget line.

Then there's scale. Running a global cloud infrastructure network requires tens of billions in annual capital investment, relationships with power utilities on six continents, and custom semiconductor design capabilities. There are exactly three companies on Earth that can do this profitably. Oracle, to its credit, has tried. They committed $50 billion in planned capital expenditure for a single fiscal year, and they still hold low-single-digit market share. That should tell you something about the height of these walls.

And finally, there's the OpenAI partnership. This one is newer and harder to value, but I think it matters enormously. Microsoft has exclusive commercial rights to integrate OpenAI's foundational models into Azure and its application layer. Every enterprise customer that wants GPT-class AI capabilities through a major cloud provider has exactly one first-party option. Competitors can offer their own models, and they do, but the integration depth between Azure and OpenAI is something that cannot be replicated through a licensing deal. You either built it together from the ground up or you didn't.

Here's where it gets interesting. These moats don't just sit next to each other. They compound. An enterprise customer using Microsoft 365 adopts Copilot AI features within the same subscription. Copilot runs on Azure. Azure runs on OpenAI models. Each layer makes the next stickier, and each dollar of revenue in one segment quietly reduces the customer acquisition cost in the others.

Coca-Cola has a flywheel. Costco has one. But I'm not sure either spins this fast with this much capital behind it.

The Regulators Are Real. And They Should Be.

I won't pretend the risks are imaginary. They're not.

The U.S. Federal Trade Commission has an active antitrust investigation examining whether Microsoft is using its dominance in productivity software to funnel customers into Azure through punitive licensing terms. The investigation is real, it is ongoing, and civil investigative demands were issued as recently as early 2026.

In Europe, the Digital Markets Act has designated Microsoft as a "gatekeeper," imposing interoperability and compliance mandates. The UK Competition and Markets Authority published a final report in July 2025 concluding that "competition is not working well" in cloud infrastructure markets, with Microsoft's licensing practices specifically cited.

These are not nuisances. If regulators forced Microsoft to make Microsoft 365 run equally well on AWS and Google Cloud as it does on Azure, that would materially damage the flywheel I just described. In a severe scenario, that kind of forced separation could compress fair value by 40% or more. I don't dismiss this.

But those of us who have been around a while remember something. Microsoft has been here before. The 1998 Department of Justice case nearly broke the company apart. Judge Thomas Penfield Jackson actually ordered a breakup. It didn't happen. The regulatory process ground on for years, resulted in a consent decree, and Microsoft emerged with its core business intact. It was a circus then, and the current investigations have a similar theatrical quality to them.

I'm not predicting the same outcome this time around. What I will say is that antitrust investigations, in practice, move slowly, settle frequently, and rarely produce the structural separations that bears build into their worst-case spreadsheets. The probability-weighted impact is real but, in my judgment, not existential.

Meanwhile, the earnings power of the underlying business compounds every quarter while the lawyers file motions. That's the part nobody prices.

A Wonderful Business. Not Yet a Wonderful Price.

Here's where discipline separates investors from speculators.

At $356.77, Microsoft trades at roughly a 10% discount to a conservatively modeled intrinsic value of approximately $395 to $406 per share. That intrinsic value assumes 14% free cash flow growth for five years, a slowdown to 10% for the following five, a terminal growth rate of 3.5%, and a discount rate of 8.5%. It accounts for the CapEx drag. It does not assume the bull case.

A 10% discount on a magnificent business is attractive. But it is not attractive *enough.*

When I buy a stock, I want a margin of safety. A gap between what the business is worth and what I'm paying that's large enough to protect me from the things I cannot foresee. For a business of this caliber, I want at least 25%. That puts my buy price at roughly $296 to $304 per share.

Microsoft would need to decline another 17% from here to reach that level. Is that possible? In this market, absolutely. The stock has already fallen 34% from its high. Tariff escalation, a broader tech selloff, or a disappointing quarter of Azure growth could easily push it there.

And if it doesn't get there? That's fine too. There are 40,000 publicly traded companies in the world. I don't need to swing at every pitch. Ted Williams became the greatest hitter in baseball by waiting for the ball to cross the plate in exactly the right spot. He let plenty of good pitches go by. So can I.

What This Means For You

If you already own Microsoft, the case for holding is strong. The business has never been more profitable, the balance sheet has never been more secure, and the competitive position has arguably never been wider. Selling because the stock is down 34% while earnings are up 16% is the precise opposite of rational behavior. You would be selling a business that is getting better because its price is getting lower. That is Mr. Market's game, and you don't have to play it.

If you don't own Microsoft and you've been waiting, keep waiting. But stay alert. A recession, a tariff war, or a single quarter of weaker Azure guidance could hand you this business at a price that would make your grandchildren wealthy. Set your limit orders. Be patient. The market rewards discipline far more reliably than it rewards urgency.

The numbers are all there in the 10-K. It's free. It's on the SEC's website. It's only 120 pages. I've read worse. I've read most of them, actually.

The fundamentals have never been stronger. The price just hasn't caught up with the fear yet. Sometimes it works the other way around. That's when the real money gets made.

---

*This narrative is not investment advice. It is an invitation to do your own homework, ideally with a calculator and the actual filings in front of you.*