AI Takes US Stock Market's Magnificent Seven Hostage: In Earnings Season, Who Is Making Money Now and Who Is Betting on the Future

As of November 2, six of the "Magnificent Seven" U.S. tech giants—all except Nvidia—had reported their latest earnings. Revenue grew almost across the board, yet their stock trajectories diverged.

The hardest fall came precisely to the fastest grower. On October 28, U.S. Eastern Time, Google parent Alphabet reported third-quarter results: Google Cloud posted quarterly revenue of $30.7 billion, up 103% year over year. Yet the next day its stock plunged 12.64%, the largest single-day drop since the company went public.

Twenty-four hours later, Amazon replayed the same scene. AWS grew 39.3%, a near five-year high, yet its shares dived more than 7% after hours. A week earlier, Meta's revenue rose 26%, and its stock also fell 7%.

The other three saw relatively stable share prices. Microsoft's Azure grew 48%, a solid result, and its stock dipped just 0.2% the next day. Tesla's operating profit fell 56%, leaving an operating margin of only 2.6%, yet because it announced it would cut full-year capital expenditure from $25 billion to $20 billion, its stock actually held steady. And Apple, which has spent almost nothing on AI infrastructure, is up about 26.7% this year, the best performer among the Seven.

The "Magnificent Seven" refers to seven U.S. tech giants: Apple, Microsoft, Nvidia, Amazon, Alphabet, Meta and Tesla. Nvidia's fiscal quarters are out of sync with the other six, and its third-quarter report will not be released until November, so this article focuses on the six that have disclosed results, while still discussing all seven when it comes to share prices and the industry chain.

In past years, the Seven's share prices moved largely in sync; this year that has changed completely. As of October 30, Apple is up 26.7% year to date, Nvidia up 9.7%, while Alphabet and Tesla are down 10.5% and 11% respectively; the seven companies account for about one-third of the S&P 500's weight, yet their weighted gain this year is only 6.2%, underperforming the broader market. This is the first time since the "Magnificent Seven" concept emerged in 2023 that they have collectively lagged.

On the surface, the market is worried about an AI bubble and whether massive spending can be recouped. In the last quarter alone, the six companies' capital expenditure combined exceeded $200 billion, a considerable share of which ultimately flowed to Nvidia—even though Nvidia's own capital expenditure for the entire previous fiscal year was only $6.1 billion. But if it were merely fear of a bubble, why is Microsoft, which is burning just as much cash, unscathed, while Google, the fastest grower, fell the hardest?

To answer that, we first need to break down these giants' businesses and see what each relies on to make money.

01. Who is earning "today's money," and who is still betting on "the future"

Microsoft and Amazon sell computing power, Meta sells advertising, Apple sells hardware and services, and Tesla bets on autonomous driving. Google is a special case, doing both: cloud and advertising. When the market starts weighing AI's returns in real money, the strengths and weaknesses of these models become apparent.

Selling computing power is currently the most stable business. As long as the whole industry is still scrambling frantically for computing power, the revenue of the three cloud providers—Microsoft, Amazon and Google—is broadly assured for the next two to three years.

Microsoft Azure posted quarterly revenue of $33.17 billion, up 48%; AWS revenue was $46 billion, up 39.3%, or $184 billion annualized, the fastest in nearly five years; Google Cloud revenue was $30.7 billion, more than doubling directly.

Google Cloud revenue doubled to $30.7 billion

Image source / Google earnings report

Whether such growth can continue depends on the orders in hand. Microsoft's undelivered commercial contracts total $677 billion, up 73% from a year earlier; Google Cloud's backlog is also $561 billion, and management said more than half will be recognized as revenue within the next two years; Amazon's backlog is $578 billion, and CEO Jassy revealed on the earnings call that OpenAI and Anthropic have each added long-term procurement contracts worth over $100 billion.

A large part of cloud providers' revenue for the coming years is already booked. The bottleneck right now is supply. Amazon currently has to install an AI chip into its data centers every 14 seconds on average, and capacity is still insufficient.

Meta takes another path: using AI to sell ads at higher prices.

Revenue this quarter was $64.6 billion, up 26%, ad impressions rose 10%, and the average price per ad rose 14%. On the earnings call, management attributed the growth to improvements in AI recommendation systems.

The problem is that the advertising business has a ceiling. Advertisers' total budgets are a finite pool, and AI's value amounts to making ads more precise and more expensive, or increasing the time users spend viewing ads. The average price already rose 14% this quarter—can it keep rising? Daily active users across all products exceed 3.6 billion—how much room for growth is left? When the marginal returns on AI investment begin to diminish, can ad growth forever outpace the growth in spending? These are all questions.

Apple has barely taken part in this arms race.

Quarterly revenue was $114.5 billion, a record, up 12%; full-year revenue was $478.8 billion, the best in the company's history. Its full-year capital expenditure was $9 billion, down 29% year over year. Over the past two years, the market has constantly criticized Apple for lagging in AI while peers splurged, and it seemed to do little. After this quarter, the wind has shifted somewhat. When everyone is fretting over massive spending and increasingly tight cash flow, "not spending" has instead become an advantage. In 2026, Apple leads the Magnificent Seven with a 27% gain.

Tesla is the fourth type: its valuation rests entirely on autonomous driving that has yet to pay off.

Quarterly revenue was $28 billion, flat with the prior year, yet operating profit fell 56%. The profit in the business of selling cars is shrinking, and the market is willing to give it a high valuation because it is betting on autonomous driving and robots. Paid FSD subscriptions reached 1.74 million, up 67%, and more than 75% of new vehicles delivered in North America were optioned with FSD; Robotaxi has also officially begun carrying passengers in Austin, with unsupervised miles growing eightfold in a single quarter.

But these are still far from making money. The potential merger that Tesla and SpaceX are evaluating also adds considerable uncertainty to its capital allocation.

Google fell the hardest, because two businesses came under question at the same time.

Search advertising revenue grew 12%, below market expectations (about 15%), and the monetization efficiency of AI search is still not as good as traditional search. Cloud revenue doubled, at the cost of $51.4 billion in quarterly capital expenditure, turning free cash flow negative at -$9.1 billion. Coupled with the drag from losses on equity investments such as SpaceX, net profit fell 24%. The earnings report came in below expectations across the board, and the market voted with its feet.

An investor who follows US stocks, Roy, told "Dingjiao One" that treating all six as a single "AI concept" to speculate on was unreasonable from the start. Even when investing in AI, some earn it back the same year, while others have to wait years; some use customers' money to build infrastructure, while others bet their own profits on the future.

02. Negative cash flow makes the market nervous

This quarter, the stock price moves of the six companies corresponded to their free cash flow situations. Free cash flow is what a company actually has left after deducting the money spent on building plants, buying equipment and buying chips.

The three with negative free cash flow—Google, Meta and Amazon—all saw their share prices fall in October. The other three with positive free cash flow—Microsoft, Tesla and Apple—all held steady.

Roy said that in the previous two years, the market was still willing to bet on "daring to spend"; now, those who spend a lot must first prove they can afford it.

The change reflects the sheer scale of this round of AI investment. Microsoft, Amazon, Google and Meta together spent $202.9 billion in capital expenditure this quarter, a single-quarter record, and their combined plans for the full year exceed $700 billion, up 75% from last year. The industry expects that scale to reach $1.3 trillion next year.

With money spent to this extent, the market naturally wants to know: when exactly will the money spent come back? Goldman Sachs summed up this shift in its industry review in the first quarter of this year: attention to AI has shifted from "who invests the most" to "how far the money invested is from turning back into cash."

The three with "negative cash flow" are not in the same situation.

Google and Amazon both spent their money on data centers, chips and networking equipment. The money spent will most likely be recovered, it just takes time. Google's capital expenditure this quarter was $51.4 billion, more than double year over year. The confidence comes from Google's $561 billion backlog, and management said more than half of it will turn into revenue within two years. Amazon's logic is the same: AWS has a backlog of $578 billion, and OpenAI and Anthropic have each added long-term purchases worth hundreds of billions of dollars.

It is worth noting that Microsoft's capital expenditure this quarter also doubled, reaching $39.4 billion, but because it earns more, operating cash flow was $58 billion, leaving $18.6 billion after deducting spending.

Meta's situation is different. Its money is also spent on data centers, with quarterly capital expenditure of $34.9 billion and a full-year plan to spend $135 billion to $145 billion, on the same scale as cloud providers. The difference is that cloud providers build data centers for customers, and customers pay once they are built; Meta builds data centers for its own use, and whether the investment can be recovered still depends on how much more advertising can earn.

In 2022, Meta made a big bet on the metaverse, and its stock fell 75% within a year; it only recovered later by laying off staff and cutting budgets. Advertising is certainly more profitable now than four years ago, but the size of this bet is several times that of back then.

More noteworthy is that Meta is shifting from "funding the AI race with its own profits" to "raising funds from the capital market": long-term debt rose from $58.7 billion to $80.9 billion within a year, and it also announced plans to create Class C shares with no voting rights. S&P recently lowered Meta's rating outlook from "stable" to "negative," citing significant uncertainty over capital expenditure and free operating cash flow.

The other two: Tesla is actively contracting, and Apple simply barely spent.

Tesla put its main resources into Robotaxi and robots, but has pulled back somewhat from earlier, larger plans: full-year capital expenditure was controlled from more than $25 billion to $20 billion, free cash flow returned to positive $1.53 billion, and it still has $30 billion in credit lines available.

Tesla quarterly operating cash flow and free cash flow

Apple barely took part in this round of investment. It does not build large-scale data centers, and its AI investment is concentrated on gradual integration into existing products and services rather than large-scale self-developed training clusters. Full-year capital expenditure was $9 billion, down 29% year over year, while quarterly operating cash flow alone was $28 billion.

But Apple also pays a price: its self-developed AI has not gone smoothly, and in January this year it announced it would spend about $1 billion a year to buy Google's Gemini model to rebuild Siri; the supply chain has also been affected, as AI servers have taken up a large amount of high-bandwidth memory capacity, the cost of ordinary memory chips has soared, and the Mac mini has already raised its price by $200.

The market reaction is actually very clear: where the repayment path is clear, the market is still willing to wait; where the path is unclear, the pressure is greatest; and those that invest conservatively, or have already begun to contract, have instead become a safe haven.

03. The rankings are being reshuffled, and the shakeout is only at halftime

Let's go back to the starting point of this round of AI. On November 30, 2022, ChatGPT was released. By the end of that year, ranked by market capitalization: Apple ($2 trillion), Microsoft ($1.8 trillion), Google ($1.1 trillion), Amazon (more than $850 billion), Tesla (less than $400 billion), Nvidia (more than $360 billion), Meta (just over $300 billion). Nvidia ranked sixth.

In May 2023, Nvidia's market capitalization reached $1 trillion for the first time during intraday trading. Two and a half years later, today, it has at one point topped the world in market capitalization.

As of October 30 this year, the new rankings are: Apple ($5.03 trillion), Nvidia ($4.93 trillion), Microsoft ($3.88 trillion), Google ($3.44 trillion), Amazon ($2.6 trillion), Tesla ($1.58 trillion), Meta ($1.57 trillion). Together, the seven have expanded from $7 trillion to $23 trillion, and their weight in the S&P 500 has risen from 22% to 33%.

Nvidia, which sells chips, has risen 13-fold, jumping from sixth place to the top, with its market capitalization at one point more than three times that of Tesla. Three years ago, the two companies had similar market values. Meta has risen nearly fivefold, but still ranks last. Along the way, Tesla also at one point fell to last place.

Image source / Unsplash

Roy said that from this quarter's earnings reports, three clues that are rewriting the rules of the game can be seen.

First, who is collecting money and who is borrowing money.

Of the huge expenditures of the six tech giants, the largest portion flows to Nvidia. Nvidia itself does not engage in large-scale infrastructure, and its capital expenditure for the entire previous fiscal year was only about $6.1 billion. As long as the industry is still scrambling for computing power, its revenue is guaranteed. But the reverse is also true: its position is essentially built on the spending of the others, and once downstream begins to contract, it will immediately come under pressure.

On the other side, some cloud vendors used to build data centers with profits; now they issue bonds, issue shares, and borrow money to build them. This quarter, Google raised a net $28.7 billion through bond issuance and also plans to issue $84.75 billion in shares, while Amazon issued $48.1 billion in bonds in a single quarter.

More macro data comes from a Reuters survey: for five tech companies—Microsoft, Alphabet, Amazon, Meta, and Oracle—operating cash flow in the first nine months of 2026 reached a record $542.9 billion, but capital expenditure doubled to $556.9 billion, resulting in a shortfall for the first time. The five companies' combined free cash flow swung from a surplus of $143.5 billion in the same period last year to a deficit of $14 billion.

Roy stressed that the market has already noticed the hidden concern of circular financing: chip manufacturers and cloud vendors use their own credit to provide loans or guarantees to downstream customers, and these funds flow back into purchases of chips and computing power, making it easy to create the illusion of booming demand.

Second, depreciation has only just begun.

What is spent on plants and chip purchases is not a one-time expense. Once equipment is purchased, depreciation begins to be accrued and amortized into the income statement quarter by quarter. Microsoft's single-quarter depreciation and amortization increased 44% year over year, and it will only accelerate going forward; Amazon's reached $79.3 billion over the past 12 months, more than its net profit of $76.5 billion in the same period a year earlier. This means that equipment is aging and depreciating faster than it is making money. JPMorgan even predicts that Amazon's free cash flow in 2027 could slide to negative $128 billion.

The reason behind this is that AI chips and servers are upgrading too quickly. A high-end GPU bought for $10 billion was expected to be amortized over ten years, but it may face the risk of asset impairment after five or even three years because its performance has fallen behind.

Investor Michael Burry recently publicly accused the giants of intentionally extending the depreciation life of AI chips, estimating that from 2026 to 2028 they may have inflated profits by a cumulative total of about $176 billion.

On this point, Apple, which has invested the least, and Tesla, which has already tightened spending, are relatively relaxed.

Third, stopping is harder than continuing to burn money.

If the commercial returns of AI materialize more slowly than expected, companies that can actively hit the brakes and adjust their pace can protect profits; otherwise, they will be dragged down by costs. Nomura's cross-asset strategist summarized this dilemma as "caught between a rock and a hard place": continuing to accelerate puts free cash flow under pressure, while slowing down undermines the AI demand narrative.

Cloud vendors have signed long-term contracts, Tesla cannot stop investing in autonomous driving, and Meta's AI recommendation system is the foundation of its advertising business and cannot stop either.

Apple is the only one with ample room to maneuver. Its investment scale is small, so it can wait until the technology route and business model are clearer before making a precise entry.

The next validation point is Nvidia's earnings report to be released in November. If its data center revenue and guidance are still accelerating, it means downstream is still ramping up; if guidance is cut, it means downstream has begun to contract.

Therefore, this earnings season has announced the end of an era: the "arms race" phase of AI infrastructure is formally giving way to the "crunching the numbers" phase. In the first half, what mattered was who was bold and who bet aggressively; in the second half, what matters is who has the strongest balance sheet and who can present investment-output results convincing enough for the market when depreciation begins to eat into profits and debt costs remain high.

This article comes from the WeChat public account "Dingjiao" (ID: dingjiaoone), author: Dingjiao One team, published by 36Kr with authorization.