Market data and company results used in this article are current through August 6, 2026.
I have a recurring problem with Microsoft stock: every time I decide it looks expensive, the company produces another quarter that makes “expensive” sound less like analysis and more like a personal failure of imagination.
This is the emotional trap of owning—or considering owning—a truly exceptional business. A mediocre company can be dismissed with a spreadsheet and a functioning sense of self-preservation. Microsoft forces me to argue with myself. The valuation looks demanding, the artificial-intelligence spending looks enormous, and the expectations look high enough to require supplemental oxygen. Then Microsoft reports 18% revenue growth on a base of more than $280 billion, Azure grows 43%, operating income rises 18%, and I am left staring at the numbers like a man who arrived at a duel carrying a strongly worded memo.
As of August 6, Microsoft shares trade around $487. The market values the company at roughly $3.63 trillion. Depending on which earnings measure and data service I use, the stock sits around the high-20s in trailing earnings. That is not a ridiculous multiple for a business of Microsoft's quality, but it is nowhere near the bargain bin. At this price, I am paying for resilience, double-digit growth, artificial-intelligence leadership, continued Azure momentum, durable enterprise pricing power, and management execution that remains almost offensively competent.
I am also paying before all of those promises have fully turned into cash.
So has Microsoft become too expensive?
The irritating answer is that the stock can be expensive and still be worth owning. The useful answer requires a real argument.
Let me make both cases.
The Numbers That Started the Argument
Microsoft closed fiscal 2026 with the kind of financial performance that makes normal corporate growth look like a hobby.
Full-year revenue reached $331.8 billion, up 18%. Operating income climbed 21% to $155.2 billion. GAAP net income rose 31% to $133.7 billion, while adjusted net income—excluding the effect of Microsoft’s OpenAI investments—grew 22% to $128.8 billion. Diluted GAAP earnings per share reached $17.95.
The fourth quarter was equally difficult for skeptics to wave away. Revenue came in at $90.0 billion, up 18%. Operating income was $40.6 billion, also up 18%. Microsoft Cloud revenue rose 27% to $59.3 billion. Azure and other cloud-services revenue increased 43%.
Azure surpassed $100 billion in annual revenue for the first time and grew 41% for the year. Microsoft 365 Copilot reached more than 30 million paid seats. Commercial remaining performance obligations—the contracted revenue waiting to be recognized—rose 84% to $678 billion.
That backlog deserves a pause.
$678 billion is not a typo produced by a frightened accountant. It is a mountain of contracted business that gives Microsoft unusual visibility into future revenue. Plenty of companies begin each quarter hoping customers will arrive. Microsoft begins with hundreds of billions of dollars in promises already standing in line.
On the other hand, the spending required to serve that demand has become spectacular. Microsoft generated $182.9 billion in operating cash flow during fiscal 2026, then spent $115.9 billion on property and equipment. Subtract the latter from the former and I get roughly $67.0 billion in conventional free cash flow.
Against a $3.63 trillion valuation, that represents a free-cash-flow yield of less than 2%.
There it is. The entire debate in two numbers.
Forty-three percent Azure growth says, “Look at the future.”
A sub-2% free-cash-flow yield says, “Yes, and look at what you are paying for it.”
The Bull Case: Microsoft Is Expensive Because It Earned the Right to Be
The strongest bull argument begins with a simple observation: very few companies combine Microsoft's scale, growth, margins, customer captivity, balance-sheet strength, and strategic position.
Investors often discuss valuation as if every dollar of earnings deserves the same multiple. It does not. Earnings from a cyclical commodity producer should not be valued like recurring revenue from software embedded in the daily operations of global enterprises. A dollar that may vanish when metal prices turn is different from a dollar attached to Microsoft 365, Azure, security, database services, GitHub, or Dynamics.
Microsoft sells products that organizations do not casually remove.
Companies can complain about licensing costs. They can hold meetings, form committees, commission studies, and appoint a vice president of strategic migration. Then somebody asks who wants to explain to 80,000 employees why Outlook, Excel, Teams, identity management, security tools, cloud infrastructure, and shared documents are all changing at once.
The migration committee is quietly dissolved.
That embedded position creates pricing power and cross-selling opportunities. Microsoft is not approaching customers as an unknown AI startup asking for a trial. It already manages their identities, documents, email, meetings, source code, cloud workloads, security controls, and business data. Adding AI to that estate is less like selling a new appliance and more like adding electricity to a building Microsoft already owns.
This distribution advantage may be the company's greatest asset.
Azure Is Still Growing Like a Much Smaller Business
Azure's 43% fourth-quarter growth is extraordinary given its scale. Azure annual revenue exceeded $100 billion, yet the business is expanding at a rate many small software companies would frame and hang above reception.
Demand has repeatedly exceeded available supply. Microsoft added 31 data centers across five continents in the fourth quarter and 88 during the fiscal year. Management says it added another gigawatt of capacity during the quarter and remains on track to roughly double overall capacity in two years.
To a bull, this is the best kind of problem. Microsoft is not building data centers in the hope that demand appears. It is racing to satisfy demand already knocking on the door.
The opportunity extends beyond renting computing power. Microsoft is building an AI stack that includes infrastructure, proprietary chips, third-party models, its own models, databases, analytics, developer tools, governance, identity, security, and workplace applications. It can make money when customers train models, run inference, store data, build agents, secure them, monitor them, and eventually ask those agents to produce a PowerPoint explaining why the agents need a larger budget.
That is a remarkably complete toll road.
The company's model catalog now includes more than 11,000 models. Foundry has 100,000 customers and revenue more than doubled year over year. Microsoft says nearly 90% of the Fortune 500 already grounds agents in enterprise context using Foundry, Fabric, and Work IQ. GitHub Copilot has reached 50 million users, while GitHub itself has 225 million.
The bull does not need Microsoft to win every layer of AI. Microsoft only needs to remain the place where enterprises buy, build, deploy, govern, and pay for it.
Copilot Is Moving From Demonstration to Distribution
For several years, the bear argument around generative AI was that usage looked impressive while monetization remained foggy. People generated poems about their refrigerators, venture capitalists described everything as transformational, and the electric grid began developing a nervous twitch.
Microsoft now has harder evidence.
Microsoft 365 Copilot has surpassed 30 million paid seats. Net seat additions more than doubled sequentially in the fourth quarter. The number of customers with more than 50,000 seats grew more than sevenfold year over year. Large organizations are deploying the product at scale: KPMG across more than 276,000 professionals, HSBC with 200,000 seats, and EY with 400,000 employees in Microsoft's largest E7 win to date.
That matters because enterprise adoption is slow by design. Large organizations do not roll software across hundreds of thousands of workers after one enthusiastic demo and a catered lunch. They test security, compliance, governance, integration, productivity gains, and whether the tool will accidentally email confidential data to a dentist in Nebraska.
Microsoft's reported adoption suggests Copilot is moving beyond experimentation.
The business model is also expanding from per-seat subscriptions toward a combination of seat and usage-based revenue. If agents become active participants in knowledge work, Microsoft may collect revenue from both access and consumption. That creates the possibility that revenue per customer rises as AI becomes more useful.
The bull case therefore is not simply “AI is big.” That phrase has financed enough questionable decisions already.
The case is that Microsoft possesses the customer relationships, infrastructure, software distribution, data access, developer ecosystem, and financial resources to turn AI from a feature into a new layer across its entire business.
The Operating Leverage Is Still Alive
A capital-intensive AI buildout should pressure profitability. In some places it has. Cloud gross margins face the weight of new infrastructure, energy, chips, and depreciation. Yet total operating income still grew faster than revenue for the full year: 21% versus 18%.
Microsoft's fiscal 2026 operating margin was nearly 47%. That is an almost indecent level of profitability for a company simultaneously spending more than $115 billion a year on property and equipment.
The Productivity and Business Processes segment generated $83.9 billion in operating income on $140.0 billion of revenue. Intelligent Cloud produced $57.0 billion in operating income on $137.8 billion of revenue. These are not speculative divisions waiting to discover a business model. They are giant cash engines financing the next infrastructure cycle from inside the company.
Microsoft also ended June with roughly $76.8 billion in cash, cash equivalents, and short-term investments, compared with about $40.3 billion in current and long-term debt. It can fund the AI race without asking capital markets for permission every six weeks.
If AI demand continues compounding, today's capital expenditures may become tomorrow's revenue capacity. Depreciation will rise, but utilization and efficiency can improve. Microsoft says it increased throughput for Copilot workloads fourfold during fiscal 2026 and is achieving better performance per dollar and per watt through its own Maia silicon and software optimization.
The bull sees the capital spending as construction, not combustion.
The Bear Case: A Great Company Can Still Be a Bad Price
Now I have to ruin the celebration.
Microsoft's quality is obvious. The stock price is also aware of it.
At roughly $487 per share, the market capitalization sits near $3.63 trillion. Using fiscal 2026 adjusted net income of $128.8 billion, the company trades around 28 times adjusted earnings. Using conventional free cash flow of roughly $67 billion, the valuation approaches 54 times free cash flow.
The earnings multiple looks reasonable only if growth remains unusually strong. The free-cash-flow multiple looks like investors have mistaken gravity for an optional software setting.
I cannot dismiss capital expenditures as temporary simply because I would prefer a prettier valuation. Microsoft is building physical infrastructure at a speed that changes the economics of the company. Property and equipment additions rose to $115.9 billion from $64.6 billion in fiscal 2025—an increase of roughly 80%.
That is more annual capital spending than the total market value of many large public companies.
Some of this investment will support long-lived assets and future revenue. Some equipment, especially advanced AI hardware, may age quickly. Chips improve, model architectures change, power requirements shift, and today's scarce accelerator can become tomorrow's expensive space heater with excellent branding.
The bear asks a question the income statement cannot yet answer: what return will Microsoft earn on this enormous new asset base?
AI Revenue Must Outrun AI Depreciation
Infrastructure spending does not hit earnings all at once. It appears on the balance sheet, then flows through the income statement gradually as depreciation. This creates a delay between writing the check and recognizing the expense.
That delay is perfectly normal accounting. It is also why I watch future margins carefully.
Microsoft's net property and equipment increased from about $205.0 billion to $313.1 billion in one year. As that larger base begins depreciating, the income statement will absorb higher costs. If AI revenue scales rapidly and utilization stays high, those costs will be spread across a large and growing revenue stream. If demand slows, customers optimize usage, or pricing falls faster than compute costs, margins could compress.
Cloud computing already teaches customers to optimize. Every chief financial officer eventually discovers the same thrilling AI use case: reducing the cloud bill.
Competition will help them.
Amazon Web Services, Google Cloud, Oracle, specialized infrastructure providers, open-source software, model developers, and customers' own hardware strategies all apply pressure. AI models may become more interchangeable. Inference prices may fall. Enterprises may use multiple clouds and route workloads toward whichever provider offers the best combination of performance, price, privacy, and negotiating desperation.
Microsoft is preparing for that world by offering many models and building its own chips. That is smart. It is also evidence that the economics are not guaranteed.
The OpenAI Relationship Is an Asset and a Complication
Microsoft's partnership and investment in OpenAI helped establish its early AI leadership. It provided access to frontier models, drove Azure consumption, and gave Copilot products a technological foundation while competitors were still trying to finish their announcement decks.
But dependence creates risk.
Microsoft is now emphasizing model choice and developing its own MAI family. Management highlighted models from OpenAI, Anthropic, Mistral, xAI, and others, while noting that models can be substituted within its broader system. Strategically, this is sensible. Financially, it tells me Microsoft does not want its future tied to one provider, one economic arrangement, or one increasingly complicated partnership.
The accounting adds noise. Fiscal 2026 GAAP net income benefited from nearly $5.0 billion in net gains related to OpenAI investments, while fiscal 2025 included losses. Microsoft wisely provides adjusted figures, but investors need to avoid treating every swing in investment value as evidence that the operating business suddenly became more or less productive.
I care more about whether AI products generate durable recurring revenue and attractive incremental returns.
The OpenAI connection may remain enormously valuable. It may also evolve as both parties develop overlapping ambitions. Partnerships between powerful technology companies tend to remain beautifully aligned until the exact moment they do not.
Microsoft's Weak Spots Have Not Disappeared
The AI and cloud story is so powerful that it can obscure weaker businesses.
More Personal Computing revenue fell 4% in the fourth quarter. Windows OEM and Devices revenue declined 7%, while Xbox content and services dropped 10%. For the full year, segment revenue was essentially flat at $54.1 billion, although operating income improved slightly.
This segment is no longer the center of the investment thesis, but it still matters. Gaming remains competitive and expensive. Hardware is uneven. Search must fight for attention in a market being reshaped by AI interfaces. Windows remains strategically important even when it is no longer the growth engine.
There are broader risks as well: cybersecurity failures, service outages, regulation, antitrust scrutiny, data-center constraints, energy availability, chip supply, and customer concern about privacy or AI reliability. Microsoft lists many of these risks itself because even trillion-dollar companies must occasionally acknowledge that the future may contain lawyers.
At a lower valuation, I can absorb more disappointment. At nearly $3.63 trillion, disappointment arrives with leverage.
The Mathematics of “Too Expensive”
The phrase “too expensive” is useless unless I define what must happen next.
Suppose Microsoft's adjusted earnings per share grows from $17.28 in fiscal 2026 at an annual rate of 15% for three years. That would produce earnings around $26.3 per share in fiscal 2029. Apply a 28-times multiple and the stock would be worth roughly $736. Discounted back at 9% annually, that future value is around $568 today, before counting dividends.
Under that scenario, $487 is not cheap, but it is defensible.
Now assume earnings grow 10% annually and the valuation settles at 24 times earnings as AI excitement matures. Fiscal 2029 earnings would reach roughly $23.0 per share, producing a future price around $552. Discount that at 9% and the present value falls near $426.
Under that scenario, the stock is already ahead of itself.
For a more optimistic case, imagine earnings compound at 20% and the market continues valuing Microsoft at 30 times earnings. Fiscal 2029 earnings would approach $29.9 per share, implying a price around $897. Discounted back at 9%, that is worth roughly $693 today.
The range is enormous because small changes in growth and valuation assumptions become very large when applied to a $3.6 trillion company.
That is why precise price targets often look more scientific than they are. I can make a spreadsheet confess almost anything if I adjust the terminal multiple with sufficient confidence.
The honest conclusion is that Microsoft's current price assumes sustained mid-teens earnings growth or something close to it. A merely good decade may not be enough. The business needs to remain exceptional.
My Bull, Base, and Bear Cases
Here is how I frame the next 12 to 18 months, using the current price near $487 as my reference point.
Bull case: $600 to $650
Azure growth remains around 35% or better, AI supply constraints ease, Copilot adoption accelerates, and investors gain confidence that capital spending is producing attractive returns. Earnings estimates rise, margins remain resilient, and the market keeps a premium multiple near 30 times forward earnings.
In this case, Microsoft's AI ecosystem becomes increasingly difficult to challenge because infrastructure, software, data, identity, and distribution reinforce one another.
Base case: $500 to $560
Azure and Microsoft 365 continue growing strongly, but capital expenditures and depreciation limit free-cash-flow expansion. Copilot adoption progresses without instantly transforming every income-statement line. The valuation remains in the high-20s as earnings grow into the current price.
This is the outcome I consider most plausible. Microsoft performs well; the stock behaves less dramatically than the business headlines suggest.
Bear case: $375 to $425
Cloud growth slows faster than expected, AI pricing becomes more competitive, capital intensity remains high, and margins begin reflecting the depreciation wave. Investors reduce the multiple toward the low-20s even if Microsoft remains profitable and strategically strong.
This scenario does not require Microsoft to fail. It requires the company to deliver results that are merely excellent when the stock price expects something closer to imperial expansion.
So, Would I Buy Microsoft Here?
I would not call Microsoft absurdly overvalued at $487. I would call it fully valued, with very little room for sloppy execution.
For an existing long-term shareholder, the business case remains strong enough that I would be reluctant to sell solely because the P/E ratio begins with a two instead of a comforting one. Selling great compounders whenever they look expensive is an efficient way to spend years watching them become more expensive without me.
For a new position, I would be more patient. I would buy gradually rather than behave as if the market will permanently abolish red days. A starter position lets me participate if Microsoft keeps compounding while preserving capital for the inevitable moment when investors rediscover anxiety.
My preferred entry would offer either a lower share price or clearer evidence that free cash flow is catching up with the infrastructure buildout. Around $430 to $450, assuming the operating thesis remains intact, the risk-reward would look more attractive to me. Below $400, I would become considerably more interested unless the decline reflected a genuine deterioration in Azure, margins, or AI demand.
The distinction matters. I want a lower price caused by market emotion, not a lower price caused by the business quietly catching fire.
What I Will Watch Next
I am watching five things.
First, Azure growth. At this valuation, a sharp deceleration would matter. I do not expect 40%-plus growth forever, but I want evidence that AI demand is expanding the cloud opportunity rather than temporarily pulling spending forward.
Second, capital expenditures. Spending can remain elevated, yet I need management to show that new capacity is translating into revenue, improved availability, and eventually stronger cash generation.
Third, cloud gross margins and depreciation. These will tell me whether AI economics are scaling or whether Microsoft is building a breathtakingly advanced machine that converts electricity into lower returns.
Fourth, Copilot adoption and usage. Paid seats matter, but sustained engagement, expansion, consumption revenue, and customer retention matter more. I want proof that enterprises see measurable value after the novelty fades.
Fifth, commercial remaining performance obligations. The $678 billion backlog is a major strength, although contract timing and large agreements can make the number lumpy. Continued growth would support the view that Microsoft has unusually deep revenue visibility.
My Verdict
Microsoft has become expensive enough that I must respect the price, but not expensive enough that I can dismiss the company.
The bull case is powerful: Azure is growing 43%, Microsoft Cloud is growing 27%, annual operating income exceeds $155 billion, Copilot adoption is moving into large enterprise deployments, and the company sits at the intersection of cloud infrastructure, workplace software, data, security, developer tools, and AI agents.
The bear case is equally real: the market capitalization is $3.63 trillion, conventional free-cash-flow yield is below 2%, annual property and equipment spending has surged to nearly $116 billion, future depreciation is climbing, competition is fierce, and the valuation assumes years of strong execution.
I do not see a bubble built on imaginary revenue. Microsoft is producing astonishing amounts of real revenue, operating income, and cash. I see a premium asset priced as a premium asset while undertaking one of the largest capital expansions in corporate history.
That makes the stock less of a screaming buy and more of a high-quality negotiation.
At $487, I would rate Microsoft a cautious hold for existing long-term investors and a gradual-buy candidate for new investors who can tolerate volatility and wait several years. I would avoid making it an oversized position at this valuation. Exceptional companies can survive almost anything except the expectations investors attach to them.
The company does not need to prove that AI matters. It has already done that.
It now needs to prove that the hundreds of billions invested in the AI era will generate returns worthy of a $3.6 trillion price tag.
I think Microsoft has a better chance of doing that than almost any company alive.
I also think the current share price knows it.
This article reflects my analysis and opinion, not personalized financial advice. Stock prices and valuation multiples change, and investors should consider their objectives, time horizon, and tolerance for loss.
Sources
Market price and valuation data as of August 6, 2026
Comments
Post a Comment