As the U.S. “Magnificent Seven” face their AI costs, who can still afford it?

Author: Chen Le | Editor: Ye Jinyan | Produced by: Shenwang · Tencent News Xiaoman Studio

After the U.S. midterm elections concluded, Wall Street's optimism toward AI plummeted within two days.

On November 5, the market capitalization of U.S. tech’s “Magnificent Seven” evaporated by approximately $1.04 trillion in a single day, but declines varied significantly among companies: Apple dipped slightly by 1.53%, while Tesla plunged 10.77%. As the biggest winner of the AI wave, NVIDIA crashed 9.26%, wiping out over $440 billion in market cap in one day, a decline 2 to 4 times larger than that of its chip buyers (Microsoft, Google, Amazon).

The trigger for this plunge was the U.S. midterm elections. After the Democrats regained control of both chambers of Congress on November 3, U.S. stocks initially rose the next day. However, by the 5th, investors shifted their focus to concerns that the new Congress might tighten AI regulations and slow down data center construction, causing the Nasdaq to fall sharply by 5.52% in a single day. On the 6th, tech stocks rebounded, with the Nasdaq rising 1.09%, NVIDIA gaining 2.78%, and AMD surging around 11%, but Apple and Microsoft continued to fall. Nevertheless, the Nasdaq still posted a weekly cumulative decline of 3.4%.

In reality, apart from Tesla, the fundamentals of these tech giants have not deteriorated: NVIDIA’s revenue doubled year-on-year, Apple set a new record high for the same period, and cloud businesses at Microsoft, Google, and Amazon along with Meta’s advertising business all maintained growth.

Economist Song Qinghui analyzed that policy uncertainty resonated with market anxiety over AI investment returns. Tech giants’ valuations have already heavily discounted AI’s high growth, yet massive investments have not yet translated into equivalent profits. In Q3 this year, Microsoft, Google, Amazon, and Meta continued to increase infrastructure spending on data centers, with capital expenditure growing much faster than operating cash flow.

Every dollar thrown by these giants becomes NVIDIA’s revenue. Once buyers run out of cash to spend, how long can NVIDIA sustain its high growth?

Chart: Capital expenditure on infrastructure by tech giants versus NVIDIA revenue trends

If Data Center Construction Halted, the "Shovel Sellers" Will Hurt Most

The answer to this question first depends on whether data centers can be built according to plan. After the election, alongside concerns about whether AI investments will yield returns, markets added another layer of worry: will regulatory policies from the new Congress and local governments slow down data center construction? Compared to cloud giants, NVIDIA, which relies heavily on downstream procurement, is more sensitive to such changes.

Chart: Pass rate of restrictive proposals for U.S. data center regulation and distribution of computing power planning

In its most recent fiscal quarter, NVIDIA’s net profit approached $60 billion, with estimates suggesting next-quarter revenue would rise by another 12% month-on-month. However, over 90% of its revenue is highly dependent on data centers. Its customer base is split evenly between cloud giants and AI startups/enterprise-government clients, with the latter growing faster. This model tightly binds NVIDIA’s performance to the pace of downstream infrastructure: faster room construction leads to heavier shipments; any progress delays result in postponed orders.

This election happened to inject uncertainty precisely at this critical juncture. On voting day, out of 41 local proposals related to data centers across 10 U.S. states, 40 leaned toward restricting data center development, with 27 restrictive or opposing measures passing. Texas's newly elected Democratic Governor Yinuo Huosa even campaigned on a promise: to fully halt data center construction until dedicated legislation is enacted.

Semiconductor research firm SemiAnalysis calculated that most data centers scheduled for delivery in 2027 have already broken ground. This election primarily impacts plans for 2028, with Texas alone accounting for nearly 30% of the U.S.'s projected new computing capacity additions for that year. Policy shifts brought by the election may mainly affect NVIDIA’s growth expectations for 2028 and beyond.

If projects are delayed, the impact on both ends of the supply chain differs: buyers can defer some capital expenditures, while NVIDIA faces the risk of delayed chip deliveries. Market movements on November 4 already reflected this divergence: according to foreign media reports, on the day investors bet on tech giants curbing AI spending, Microsoft and Amazon bucked the trend, rising roughly 6%, while NVIDIA fell alone.

By the 5th, the sell-off spread across the entire AI supply chain. Amid dual concerns over weak Qualcomm earnings and the market's fear that policies would slow AI infrastructure rollout, the Philadelphia Semiconductors Index plummeted 12% in a single day, a decline far exceeding that of cloud computing companies.

But this does not mean computing demand has peaked. Song Qinghui cautioned that algorithmic trading, leveraged capital, and options hedging significantly amplify short-term volatility, meaning a single-day crash does not imply a proportional loss in intrinsic corporate value. On the 6th, NVIDIA rebounded 2.78% but had not yet recovered the 9.26% drop from the previous day. The upcoming quarterly earnings report on November 17 will serve as a crucial window for the market to assess its growth prospects.

In contrast, Apple and Tesla sit at opposite ends of this logic chain. Song Qinghui pointed out that their growth models differ fundamentally from those of AI infrastructure companies: Apple’s core profitability lies in smartphones and services, with lower reliance on data center construction, making it relatively resilient; Tesla suffered the steepest decline, exacerbated by Musk’s political contributions. Having invested over $400 million previously to support the Republican Party, Musk’s business empire could face heightened regulatory scrutiny following the Democratic victory. Within two days after the election, the combined market capitalization of Tesla and SpaceX evaporated by more than $400 billion.

Who Will Foot the Next Massive Bill?

Whether server rooms can ultimately get approved comes down to financing. This massive funding is primarily shouldered by NVIDIA’s core buyers: Microsoft, Google, Amazon, and Meta. Latest earnings reports show that all four are still increasing AI investments, but they can no longer rely solely on internal cash flow to cover the costs.

Chart: Comparison of operating cash flow versus capital expenditure gaps for Microsoft, Google, Amazon, and Meta

A year ago, the combined operating cash flow of these four companies exceeded fixed asset expenditures by approximately $61.7 billion; by Q3 this year, that surplus had turned into a deficit of roughly $26.4 billion.

This is not because AI isn’t profitable. Google Cloud’s Q3 operating profit tripled year-on-year, AWS operating profit rose from $11.4 billion to $17.8 billion, and Microsoft Azure revenue grew by 48%.

The shortfall stems from a timing mismatch: money for building data centers must be paid upfront, while revenues take years to realize. Take Google as an example: it holds $561 billion in cloud contracts, over 55% of which require recognition over a two-year period. Yet its signed future procurement commitments reach $924 billion—1.65 times the contract value.

More and more of these upfront payments are being financed through borrowing. Within a year, Google increased its long-term debt from approximately $20 billion to $125 billion and launched a stock offering plan exceeding $80 billion. Over the past 12 months, Amazon’s free cash flow shifted from a $14.8 billion inflow to a $42.4 billion outflow, largely covered by bond issuance. For Meta, free cash flow turned negative for the first time in over a decade this quarter, with its long-term bonds trading at levels comparable to many BBB-rated corporations.

Prudential Investment believes this reflects investor concerns over excessive spending and uncertain AI monetization. Among the four, only Microsoft still boasts approximately $18.6 billion in free cash flow and has avoided large-scale debt issuance in recent years. According to Bank of America statistics, the top five cloud providers issued about $121 billion in USD-denominated bonds collectively in 2025, more than quadruple the average annual level of the preceding five years.

As external financing scales up, financing costs are beginning to significantly influence the pace of data center expansion. Song Qinghui noted that persistent inflation and high interest rates will drive up borrowing costs, and projects relying on external financing with excessively long payback periods should appropriately slow their expansion. Currently, the 10-year Treasury yield remains above 4.8%. Approval processes pose another hurdle. SemiAnalysis points out that while tech giants can afford the costs imposed by new regulations, it does not guarantee timely computing power delivery, and projects halted by injunctions struggle to quickly relocate to other states.

Under the most pressure are NVIDIA’s fastest-growing customers. Pan Helin, member of the Expert Committee on Information and Communication Economy under China’s Ministry of Industry and Information Technology, noted that unlisted large model firms like OpenAI and Anthropic heavily rely on external financing. If interest rates remain stubbornly high and regulations tighten, the difficulty of securing subsequent funding will surge dramatically.

This is exactly why NVIDIA fell harder than its buyers. Nigel Green, CEO of deVere Group, argues that investors will now scrutinize each company individually to see who can truly convert AI investments into revenue.

On November 6, despite a rebound, NVIDIA had not yet recovered most of the previous day’s losses. Compared to Apple’s profit model, which is relatively independent of AI infrastructure build-outs, NVIDIA’s high growth relies more heavily on sustained downstream investment. The speed at which buyers can continue constructing data centers will directly dictate the next growth phase for this "shovel seller."