1. Bill Ackman's case for the AI hyperscalers
Bill Ackman, who runs Pershing Square Capital Management, says Wall Street is mispricing Amazon, Microsoft, Meta Platforms and Alphabet. In the firm's third-quarter 2026 shareholder letter, he argues analyst forecasts fully absorb the roughly $740 billion these four companies are expected to spend on capital projects in 2026, yet give little credit to the revenue that AI infrastructure could produce.
The four companies are commonly grouped as the hyperscalers: operators of cloud computing and data center networks at a scale few others can match. They are also the largest buyers of the chips, servers and networking gear that power artificial intelligence, which places them at the center of the current investment cycle.
Ackman's argument is that the market is looking at only one side of the ledger. Analysts, in his view, are modeling the enormous cost of building AI capacity as if it were a permanent drag, without assigning much value to the earnings that capacity could generate once it is running and sold to customers.
Shareholder letters of this kind are a long-standing tradition among activist and concentrated investors. They explain the reasoning behind a fund's largest positions and often serve as a public argument for why the market has misjudged a company. Pershing Square typically holds a small number of large stakes, so its views on any single holding carry significant weight for the fund's overall performance.
2. A capital spending bill that has grown five-fold
The scale of the spending is the starting point for the debate. The letter puts the combined capital expenditures of the four companies at about $140 billion in 2023. Based on company guidance, that figure is expected to reach roughly $740 billion in 2026, an increase of more than five times in three years.
Consensus estimates cited in the letter point to a further rise of around 35% next year, which would lift the group's combined spending above $1 trillion in 2027. Few corporate investment programs in history have grown this quickly, and the numbers have understandably made many investors uneasy.
Capital expenditures, often shortened to capex, are the money a company spends on long-lived physical assets such as buildings, land, equipment and computing hardware. Unlike operating costs, capex is spread across many years on the income statement through depreciation, but it reduces free cash flow immediately. A sudden jump in capex therefore tends to compress the cash available for dividends, buybacks and other uses in the near term, even if the investment pays off later.
That timing gap is at the heart of the market's caution. Investors are being asked to accept a sharp drop in near-term cash generation in exchange for revenue that has not yet fully arrived, and some have responded by trimming their exposure to the companies doing the spending.
3. Offensive rather than defensive spending
Ackman frames the question in simple terms. Investors, he notes, have been asking whether the hyperscalers are spending defensively, simply to hold on to their existing businesses, or offensively, to build new sources of growth that will earn strong returns in the future.
His answer is clear. The letter describes the spending as offensive rather than defensive, arguing that the companies are not merely protecting their franchises but investing to accelerate growth. In this reading, the hyperscalers are spending because they see a large new market to capture, not because they are being forced to keep up with rivals.
The distinction matters for valuation. Defensive spending is closer to a maintenance cost: necessary, but unlikely to lift profits. Offensive spending, if it works, should show up in faster revenue growth and higher earnings over time, which would justify a higher share price than a purely defensive reading would suggest.
4. Earnings up 27%, share prices up 8%
The letter highlights a striking gap between business results and stock market returns. Earnings per share at the four hyperscalers have grown by about 27% this year. Their share prices, however, have risen by only about 8% on average year to date through early November 2026.
The suppliers of the AI build-out have done far better. The S&P 500 Semiconductor Index has gained 44% over the same period, and the S&P 500 Technology Hardware Index is up 36%. Pershing Square treats these indexes as rough proxies for the picks-and-shovels companies that sell chips and equipment to the hyperscalers.
Put another way, the market has rewarded the companies receiving the AI spending far more generously than the companies doing it. When earnings grow faster than share prices, the price-to-earnings multiple shrinks, and the letter attributes much of that compression to investor worries about the returns the hyperscalers will earn on their elevated capex.
For a value-oriented investor, that kind of divergence is often where opportunities are found. If the earnings growth proves durable and the spending begins to generate visible revenue, the multiple could expand again, delivering returns from both higher profits and a higher valuation.
The contrast also reflects how the market has chosen to express its enthusiasm for AI. Rather than betting on the companies that must eventually turn computing power into products and services, many investors have preferred the suppliers whose sales rise directly with every new order for chips and equipment. That preference is rational in the early stages of a build-out, when supplier revenue is immediate and buyer revenue is still developing, but it can shift once the buyers begin to show returns.
5. How much can a gigawatt of AI capacity earn?
Pershing Square estimates that each gigawatt of new AI data center capacity costs roughly $40 billion to $50 billion to build, and that a hyperscaler can generate about $15 billion to $20 billion in annual revenue from it. Those numbers form the core of the firm's case that the build-out can pay for itself.
Data center capacity is increasingly measured in gigawatts of power rather than in square feet or server counts, because electricity has become the binding constraint on how much AI computing a site can support. A single gigawatt is enough to run an enormous number of high-performance accelerators, and building that capacity involves land, power infrastructure, cooling systems, buildings and the chips themselves.
If a facility costing $40 billion to $50 billion can bring in $15 billion to $20 billion every year, the investment could be recovered over a period of a few years, depending on margins and operating costs. That is the arithmetic behind Ackman's view that the market is too focused on the upfront bill and not focused enough on what follows.
6. A 5 to 10 point revenue beat through 2029
The letter's boldest projection follows from those economics. If the growth portion of the hyperscalers' capex earns returns in line with the one-gigawatt example, Pershing Square estimates their revenue growth could exceed consensus forecasts by five to ten percentage points a year through 2029.
That would be a very large gap. Consensus estimates represent the average expectations of analysts covering each company, and they heavily influence how the stocks are valued. An annual beat of that size, sustained over several years, would compound into substantially higher revenue and earnings than the market currently assumes.
The firm's broader point is that current forecasts embed the cost side of the build-out while assuming little acceleration in sales. In Pershing Square's view, that combination penalizes the companies for spending today without recognizing what the spending is likely to deliver.
7. Where Pershing Square has put its money
Ackman has backed the thesis with capital, though not across all four names. Pershing Square has bought shares of Microsoft, Meta and Amazon during periods when the stocks fell on concerns about their capital spending.
Discussing the Meta position in a television interview, Ackman said the fund was able to buy at a steep discount after Meta unveiled its capital spending plans and the stock sold off. That approach, buying when the market reacts negatively to a spending announcement, is consistent with the letter's view that investors are overreacting to the cost side of the story.
Alphabet is the exception. Pershing Square sold its stake in the Google parent earlier in 2026, which means the fund currently owns three of the four companies its letter defends. Readers weighing the argument should keep in mind that it supports positions the fund already holds.
8. What it means for investors in Big Tech
If Ackman's reading proves correct, the gap between roughly 27% earnings growth and roughly 8% share price gains could narrow as AI revenue becomes more visible in quarterly results. Investors who bought during capex-driven sell-offs, as Pershing Square has done with Microsoft, Meta and Amazon, would stand to benefit most.
The risks are equally clear. The revenue projections depend on pricing, demand and utilization of AI capacity holding up for years. If customers pay less for AI services than expected, if demand grows more slowly, or if new capacity sits partly idle, the returns would fall short of the letter's assumptions.
The sheer size of the commitment also reduces the margin for error. With combined spending on a path toward more than $1 trillion a year, even modest disappointments in revenue could translate into large shortfalls in returns on capital. That is the scenario the market appears to be pricing in today, and it is the scenario Ackman believes is too pessimistic.
For now, the debate is likely to be settled quarter by quarter. Each earnings report from the four companies will offer fresh evidence on whether AI revenue is catching up with AI spending, and on whether Wall Street's caution or Pershing Square's optimism is closer to the mark.