Why Ackman Bet $2.4 Billion That the Market Misunderstood Microsoft
There is a difference between buying a stock because it is going up and buying a stock because the market does not yet know what it is worth. Bill Ackman, founder of Pershing Square Capital Management, built a position of $2.4 billion in Microsoft by doing exactly the latter. The distinction is not semantic: it defines who takes on structural risk and who is simply riding a trend.
In February 2026, Microsoft published its second quarter fiscal results. The market reacted with concern over the pace of growth in the cloud business and over a capital expenditure plan that, viewed from the outside, appeared aggressive relative to visible revenues. The stock fell. Ackman started buying. By the time the position was documented in regulatory filings, Pershing Square had accumulated between 5.6 and 5.7 million shares, equivalent to 15% of the fund's equity portfolio, a concentration that in a high-conviction fund of that caliber is not an accident but a declaration.
The entry price, according to Ackman, was around 21 times forward estimated earnings. For a company with Microsoft's competitive profile, that multiple is strikingly low. The company has historically traded at a premium to the broader market, and at the time of purchase it was trading roughly in line with it. That, in terms of risk structure, means Ackman did not pay a premium for quality: he obtained it without paying extra.
The question that should guide the analysis is not whether the bet is good or bad. It is whether the internal logic of the case holds up to scrutiny once the conditions that made it possible are removed.
The Argument Hidden Beneath the Artificial Intelligence Narrative
When an investor of Ackman's stature explains his thesis publicly, you must read both what he says and what he does not say but what underpins the position. In this case, the visible narrative is artificial intelligence and Azure. But the more interesting structural argument is a different one: that the market is undervaluing an asset within Microsoft that is not yet properly reflected in conventional valuation models.
That asset is the approximately 27% stake that Microsoft holds in OpenAI. Ackman estimated that this stake, valued independently, could represent approximately $200 billion, which would be equivalent to 7% of Microsoft's market capitalization. If that calculation has any basis, it means that at Pershing Square's entry price, the market was effectively giving away a significant fraction of that asset when valuing the company.
That changes the nature of the bet. It is not merely a wager on Azure's growth, which is already a widely covered and consensus-discounted thesis. It is a bet on a specific valuation asymmetry: the market is using an inadequate model to value Microsoft because it is systematically ignoring or underestimating a component of the balance sheet that does not fit neatly into traditional analytical frameworks.
Pershing Square is not the only entity capable of performing that calculation. The difference is that Ackman took the position when short-term noise surrounding capital expenditure created a window to enter at a low multiple. That is what distinguishes a structurally coherent move from a merely opportunistic one.
What reinforces the strength of the argument is the context of the financing. Pershing Square built the position in Microsoft by reducing or exiting its stake in Alphabet. Ackman was explicit in stating that the sale of Alphabet was not a bet against that company, but a reallocation of capital toward where he saw greater asymmetry. That reveals the underlying mechanics: it is not a broad expansion of aggregate exposure to the technology sector, but a redistribution within it, built around a fairly precise hypothesis of relative valuation.
The Risk That Does Not Disappear When the Story Works
Building the logic of the case is not equivalent to eliminating the risk. And in the case of Microsoft, the risks have structure, not just narrative.
The first and most significant is the level of committed capital expenditure. The company has projected spending approximately $190 billion on artificial intelligence infrastructure and cloud computing during 2026. That figure is not discretionary in the sense that it can be halted without consequences: it implies contracts, physical infrastructure, supplier relationships, and commitments to customers who depend on that capacity. If demand for artificial intelligence services does not grow at the pace that justifies that capital deployment, or if the pricing of those services compresses before the capacity generates returns, the penalty on free cash flows could be severe and prolonged.
The second risk is competitive in origin. Azure competes directly with Amazon Web Services and Google Cloud in a market where all three platforms have comparable economies of scale, rapidly developing artificial intelligence capabilities, and access to the same large corporate clients. The thesis that Microsoft holds a durable structural advantage in enterprise artificial intelligence may be correct, but it depends on that advantage materializing in customer retention, contract expansion, and pricing power. None of those three factors is guaranteed by the current position in the market.
The third risk is the most difficult to manage: the fragility in the valuation of OpenAI. The $200 billion estimate that Ackman places on that stake is not an audited figure nor an observable market price. It is an inference based on the implied valuation from private capital funding rounds and on projections about OpenAI's capacity to monetize its technological position. If OpenAI encounters competitive, regulatory, or business model turbulence, that number could be revised dramatically downward, and with it, the portion of Ackman's argument that depends most heavily on a non-conventional valuation.
That does not invalidate the thesis, but it compels precision about which part of the case is solid and which is most exposed to revision.
What the 15% Concentration Reveals
A fund with Pershing Square's philosophy does not arrive at 15% of portfolio in a single position through gradual accumulation without conviction. That concentration is a signal of risk reading, not merely enthusiasm. To understand why, one must examine how the logic of a concentrated fund operates.
When a high-conviction manager places 15% of their portfolio in a single position, they are implicitly making two claims. The first: that the range of probable outcomes in that position is sufficiently asymmetric to justify the concentration — that is, that the loss scenarios are bounded and the gain scenarios have sufficient breadth. The second: that they possess better information, a better analytical framework, or both, relative to the consensus that set the price at which they were able to buy.
In the case of Microsoft, both claims have a basis. The scenario of serious structural loss — meaning that Microsoft loses competitive relevance in cloud and artificial intelligence — implies a sequence of events sufficiently complex and prolonged that an investor with a long time horizon could react before the damage becomes irreparable. And the thesis of the undervaluation of the OpenAI asset, if correct, represents a revaluation catalyst that the consensus was not adequately integrating in February 2026.
The problem with concentration is not the risk itself. Concentrated funds manage risk differently from diversified ones, not necessarily worse. The real problem emerges when the concentration is based on an argument that has a critical dependency: in this case, the valuation of OpenAI. If that argument fails, the weight of the 15% makes the correction more painful than it would be if the fund had a distributed exposure.
For an individual investor who considers Microsoft a relevant position in a diversified portfolio, the risk structure is different. They do not have the continuous analytical capacity of a hedge fund of this caliber to monitor the specific catalysts of the case, nor to exit at the right moment if the thesis begins to deteriorate. That does not mean the position is wrong — it means that the appropriate size is structurally different.
The Asymmetry That Remained After the Rebound
The question at the heart of the original article — whether Microsoft remains a buy after the rebound — is the right question, but poorly framed. There is no universal answer because the answer depends on which part of Ackman's thesis one shares and at what price one is entering today relative to the price at which he entered.
At 21 times forward estimated earnings, the risk-and-return equation had a specific shape: low multiple, concrete catalysts, favorable asymmetry. After the rebound, the multiple has risen, the catalysts are now more widely known by the market, and part of the initial asymmetry has been captured. That does not turn Microsoft into a structurally weak position, but it does mean that the margin of safety available to Ackman in February is not available at the current price.
What remains intact, if the thesis is correct, is the long-term component: Azure's infrastructure as a platform for enterprise artificial intelligence growth and the stake in OpenAI as an asset that has not been fully valued. Those two elements do not depend on the entry price, but on time and on the operational fundamentals behaving in accordance with the hypothesis.
The central risk — the one that defines the structural quality of this bet even after the rebound — is not the current multiple nor the competition in cloud. It is the dependence that the valuation argument places on a private, illiquid asset subject to considerable regulatory and competitive uncertainty. That dependence is the point of fragility that any investor who decides to follow Ackman's thesis must have explicitly in their model, not as a footnote but as a central variable. A bet of this magnitude on a non-observable asset is not inherently weak, but it demands that whoever replicates it understands that they are accepting a source of uncertainty that Microsoft's public numbers alone will not resolve.
