When Cash Burning Is No Longer a Valid Pass: Microsoft and Meta Hit the Brakes, Apple Emerges as the AI Winner

Microsoft and Meta are adjusting their AI spending. Investors have discussed Anthropic's IPO valuation expectations from around $2 trillion to around $1.5 trillion, while Apple — the company once widely considered to have fallen completely behind in the AI race — has been repriced as a winner precisely because it "never truly joined the LLM race".

Viewed in isolation, each of these developments can be interpreted as an unrelated event. But taken together, they point to the same signal: the core of the AI narrative is shifting from "who dares to spend" to "who can recoup the money they spent".

The "cuts" by Microsoft and Meta are not retreats, but accounting adjustments

Let's start with Microsoft. A channel survey by TD Cowen shows that Microsoft has canceled data center leases with a total capacity of "several hundred megawatts" in the United States, and suspended the process of converting signed letters of intent into formal lease agreements. Analysts conclude that Microsoft "foresees oversupply".

Microsoft's response is very nuanced: its $80 billion infrastructure investment plan for this fiscal year is "still proceeding as scheduled", but it "may adjust the strategic pace or adjust infrastructure in certain areas". At the same time, Microsoft has lowered its capital expenditure guidance for the 2026 calendar year from about $1900 billion to about $1750 billion, partly due to an accounting adjustment for the service life of data centers — extending the depreciation period from 15 years to 25 years, and reclassifying a large number of assets from finance leases to operating leases. (These two figures belong to different fiscal years and accounting calibers, and are not suitable for direct comparison.)

The situation at Meta is similar. In 2026, Meta's capital expenditure guidance ranges from about $125 billion to $145 billion (the range has narrowed at some points in time), a sharp increase from $72.2 billion in 2025. However, in the first half of this year, Meta launched layoffs involving about 8,000 positions to "hedge the pressure of AI-related spending". Goldman Sachs interprets this as "making room for new growth" rather than "cutting budgets to survive the winter" — the released resources will be reallocated to AI growth investments centered on computing power demand.

Essentially, the moves of these two companies do not mean "stopping investing in AI". They are investing more than ever before. The change is that they have begun to conduct stricter reviews of the pace, structure and return period of AI investments. Microsoft extending the depreciation period of data centers and Meta using layoffs to hedge infrastructure costs are both patching up the "spare no cost to seize computing power" model of the past two years. Steffen Van Nieuwerburgh, a professor at Columbia Business School, pointed out that in 2026, the combined capital expenditure of Microsoft, Amazon, Meta, Oracle and Alphabet will for the first time exceed their total operating cash flow. When the growth rate of capital expenditure exceeds the growth rate of operating cash flow, the pressure of tightening financing environment will change from "expectation" to "constraint".

The downward adjustment of Anthropic's valuation reflects that the capital market is repricing "growth". Anthropic's Investor Day delivered a more direct signal. On October 14 in San Francisco, investors discussed a valuation of about $1.5 trillion, lower than the previous expectation of about $2 trillion for the IPO scheduled in November.

The direct cause is the slowdown in revenue growth. Anthropic's revenue in the third quarter was $16.8 billion, a month-on-month increase of about 45%, and it was profitable on an adjusted basis. For the vast majority of companies, these are strong figures. But in the second quarter, its revenue more than doubled from $4.7 billion in the first quarter to $11.6 billion. The slope of the growth rate changed from "more than doubling" to "about 45%".

Investors' reaction illustrates one thing: for cutting-edge AI companies, "just growing fast" is no longer enough. What the market needs to judge is whether the speed of revenue growth can keep up with the speed of computing power investment. Anthropic has previously promised to invest $518 billion in computing power over the next ten years, while its current annualized revenue run rate is about $67 billion. This gap requires sustained high-speed growth to fill. When signs of slowing growth appear, valuation multiples will be re-examined.

This IPO is regarded by the market as a recent weathervane for the entire AI sector, and OpenAI's potential listing next year is also under observation. The pricing result of Anthropic will directly affect the public market's valuation framework for the asset class of "AI labs".

Apple's "win" lies in that it does not need to burn money on AI

The logic behind Apple's repricing is completely different from that of Microsoft, Meta and Anthropic.

In July 2026, Apple's market value briefly touched $5 trillion, making it the second company in history to reach this milestone. Analysts expect Apple's AI-related expenditure in 2026 to be about $11 billion to $13 billion, while Microsoft, Google, Amazon and Meta each have capital expenditure at the level of hundreds of billions of dollars.

Apple's path is: instead of building large-scale data centers or developing cutting-edge foundational models on its own, it embeds AI capabilities into iPhone, Mac, iPad and other end devices through cooperation with partners such as Google. Its moat does not lie in the scale of computing power, but in the terminal distribution channels of billions of active devices and the closed loop of user data.

This strategy was widely criticized in the past two years as "falling behind in the AI race". But when investors began to worry about the return period of AI capital expenditure, the debt-financed computing power construction, and whether model companies' revenue growth can match the scale of investment, Apple's "no money-burning" feature has turned from a liability into an asset. Apple's stock price has risen by about 25% in 2026, the best performance among the "Magnificent Seven Tech Giants", while Tesla, Microsoft and Meta have all fallen over the same period.

This is not because Apple's AI products are more outstanding — its AI feature rollout has been stumbling all along. Instead, in a race that requires continuous huge capital investment and has an unclear return path, not participating in the race itself has become a way of risk aversion.

The common direction of these three events

Putting Microsoft, Meta, Anthropic and Apple together, they describe different sides of the same thing.

In the past two years, the main line of the AI narrative was "investment equals faith". As long as companies were willing to spend money on GPUs and data centers, the capital market was willing to assign high valuations. The capital expenditure curves of Microsoft, Meta, Google and Amazon kept rising, and the valuations of model companies jumped by units of tens of billions of dollars.

Now, the constraints of this main line have begun to become explicit. Cash flow is limited, financing costs are rising, and revenue growth can't double forever. The adjustments of Microsoft and Meta are to re-find a balance between capital expenditure and operating cash flow. The downward valuation of Anthropic means that the public market is demanding "growth quality" rather than just "growth speed". Apple's repricing reflects that capital is rotating from the "computing power narrative" to the "cash flow narrative".

This does not mean that the AI story is over. Data from JPMorgan Chase shows that in 2026, the capital expenditure of major cloud vendors will still be close to $1 trillion, a year-on-year increase of 97%, but the growth rate in 2027 is expected to slow down to 47%. The absolute scale of investment is still staggering, and what has changed is the marginal growth rate and the market's patience threshold.

What the AI industry is experiencing is not the bursting of a bubble, but a stage shift from "narrative-driven" to "financial discipline-driven". In this transition period, the first companies to be rewarded are no longer those that shout the largest investment figures, but those that can draw a clear path between investment and return — or, like Apple, those that do not need to draw such a path at all.

This article is from the WeChat Official Account "Tech Business", authored by Tech Business, and published with authorization from 36Kr.