After the Nasdaq suffered a two-week slump, "faith in AI" is teetering! Capital is shifting from chasing growth to questioning the return on AI investment.

The core logic behind this round of global tech stock repricing can be described as Wall Street asset management giants simultaneously beginning to scrutinize the actual scale of AI-driven growth, the cost of capital expenditures, and the timeline for realizing returns on AI-application-related investments. This week, tech companies including Google parent Alphabet, Microsoft, and Amazonwith a combined market value exceeding $10 trillionwill report earnings. All three are expected to achieve double-digit overall revenue growth, but what investors care about most is undoubtedly how much more AI capex will increase, and when the newly added AI computing infrastructure investment will generate sufficiently strong AI cash returns. It is understood that the recent sharp selloffs across global financial markets in AI computing-themed stocks, broad tech stocks, and AI-related bond assets have fully highlighted that the market has already expressed unease and anxiety in advance: as of the close on Friday, October 23, the Nasdaq 100 Index had fallen about 6.24% over two weeks, its worst two-week performance since April 2025; in particular, on October 22, the Nasdaq Composite fell 2.60%, and the Philadelphia Semiconductor Index plunged 6.72%, with the latter rebounding only 0.44% the next day, tracking the modest rebound in Korean and Chinese tech stocks. The prolonged and recently escalating energy-driven inflation shock has further raised the market's required return on investment, specifically reflected in the surge in yields on 10-year and longer-dated U.S. Treasuries. The Middle East war, risks to energy facilities and shipping, and the Russia-Ukraine conflict's disruption to energy supplies continue to push up inflationary pressure. The pricing benchmark for the international crude oil marketBrent crudesettled at $116.72 per barrel on October 23, the highest close since April 29, up about 61% from $72.48 on the last trading day before the Iran war broke out. U.S. gasoline prices rose 9.1% month-over-month in September, contributing more than half of the overall increase in consumer prices that month. Although safe-haven buying has pushed the 10-year Treasury yield down from its early-October high to around 5%, tech companies associated with AI computing infrastructure still face a relatively high benchmark cost of AI financing. For new cloud providers that rely on external capital to build computing power, the pressure comes from the persistent combination of high benchmark long-dated Treasury yields and widening credit spreads. Market pricing on October 22 showed that the yield on a CoreWeave bond maturing in 2031 reached about 15.76%, with five-year CDS at about 1,115 basis points; pricing reported by media on October 15 showed that CoreWeave's five-year CDS corresponded to an implied five-year cumulative default probability of about 63%, compared with about 30% a year earlier. This probability incorporates recovery-rate assumptions and risk premiums, reflecting the risk compensation the market demands. Japan's investment giant SoftBank Group, which has invested heavily in OpenAI in recent years and has bet almost "all-in" on artificial intelligence, previously issued three tranches of dollar bonds in September with coupons of 8.625%, 9.25%, and 9.75%, and its five-year CDS also surged to about 432 basis points on October 20 due to the surge in AI financing costs. The build-first, get-paid-later model is bearing more expensive funding costs. Large tech giants such as Microsoft, Amazon, and Meta, with long-strong fundamentals and long-mature growth in their core businesses, are also affected by credit repricing, but cash flow from mature businesses gives them stronger financing capacity. Meta's five-year CDS was about 123 basis points on October 23, while Amazon's related metric had previously approached its highest level since March 2020. In addition, regarding the cash-to-debt metrics of tech giants, verifiable quotes on October 5 showed that the yields on Microsoft bonds maturing in 2060, Amazon bonds maturing in 2076, and Meta bonds maturing in 2064 were about 6.36%, 6.95%, and 7.48%, respectively. Quotes for bonds of different maturities reflect investors' required compensation for long-term funding and credit risk, and do not mean that the coupons on existing fixed-rate debt will rise immediately. What is truly noteworthy now is that a modest pullback in Treasury yields and widening credit spreads for AI companies can occur simultaneously, so that companies' new financing costs may not decline in tandem with the 10-year Treasury yield trajectory. AI-spending-related worries may overshadow the tech giants' strong results Among the world's largest and most important tech giants, three will report earnings and future outlooks this week, and Wall Street analysts almost unanimously expect them to show healthy growth trajectories. Normally, this would be seen as a catalyst to stabilize a turbulent stock market. But this time, for Wall Street analysts and hedge fund traders, everything depends on two data points: how much money they are spending on artificial intelligence, and when those investments will begin to generate substantive returns. Judging from last week's sharp market reaction to increased capital expenditures at Facebook and Instagram parent Meta Platforms, any disappointing performance could be treated harshly. John Belton, a senior portfolio manager at Gabelli Funds, said investors are currently "completely out of patience with further increases in capital expenditures." "The stock market can turn a blind eye to risks for a while, but those risks will eventually have an impact, at least making the market more fragile." With oil prices soaring and Brent crude rising to its highest level since April, investors were already highly anxious heading into this week. Traders also rushed back into relatively safe U.S. Treasuries, pushing the 10-year yield below 5% for the first time in a month. Combined with the ongoing Iran war and next week's critical U.S. midterm elections, the macro environment is full of challenges. As shown in the chart above, this week brings earnings from tech companies with a combined market value exceeding $10 trillionover the next five days, about 36% of S&P 500 constituent companies will report results. Against this backdrop, tech giants including Alphabet, Microsoft, and Amazonamong the biggest AI spenderswill report earnings one after another. All three companies are expected to achieve double-digit percentage revenue growth for the quarter. But what investors care about most is how much they plan to invest in AI, and to what extent those investments have proven profitable. Kevin Shea, senior equity analyst at BNY Wealth Management, said: "This is a fairly critical earnings season. The market is discounting the growth prospects of mega-cap companies because it is increasingly uncertain whether that growth can be sustained." Meta's stock performance after reporting earnings on Wednesday afternoon clearly reflected this unease. The stock surged in after-hours trading as results showed encouraging trends. But by Thursday morning's open, sentiment had reversed as investors digested the news that the company planned to increase future spending commitments by $125 billion to $825 billion. Meta closed down 6.1% on Thursday, its worst single-day performance since July; the price of its five-year credit default swaps also rose to a new high, as the massive spending means the company will need to issue more debt. In recent weeks, multiple concerns in the tech industry have intertwined, driving a rapid shift in investor attitudes toward AI spending. First, comments from cloud service provider CoreWeave deepened market worries about the extent to which AI infrastructure buildout relies on debt financing, especially as interest rates continue to rise and the risk of project problems amplifies accordingly. Research firm Gartner estimates that if AI revenue cannot justify the spending, more than $3.5 trillion in AI infrastructure investment could face impairment risk by 2030. Jed Ellerbroek, portfolio manager at Argent Capital Management, said: "We are currently in an extreme risk-off state. Investors are fleeing stocks related to data center capital expenditures, fleeing tech and data-center-related assets." Subsequently, Anthropic, a large AI lab expected to list next month, disclosed to potential investors that its third-quarter revenue grew about 45% year-over-year, slower than in the second quarter, casting a shadow over the industry's growth narrative. Gabelli's Belton said: "A tough macro environment, slowing AI revenue growth, and accelerating AI capital expendituresthat's a tricky combination." For years, mega-cap tech stocks have been the main force driving the market higher, but that changed in 2026. The Bloomberg "Magnificent Seven" index has risen only 6.5% year-to-date, notably lagging the S&P 500's 11% gain and the 15% gain of the tech-heavy Nasdaq 100. Among the seven companies, only Apple and Nvidia have outperformed the market. The S&P 500 excluding these tech giants is up 12% year-to-date. As shown in the chart above, the U.S. stock market's "Magnificent Seven" (Mag Seven) stocks have significantly underperformed the broad U.S. market benchmark year-to-date. The so-called "Magnificent Seven" (Mag 7), which hold a large weight (over 40%) in the S&P 500 and Nasdaq 100namely Nvidia, Apple, Microsoft, Google, Tesla, Amazon, and Facebook parent Meta Platformsare the core drivers of the S&P 500's repeated record highs and are regarded by top Wall Street investment institutions as the portfolio most capable of delivering huge returns to investors amid the largest technological transformation since the internet era. Investors appear to hold two intertwined views on AI spending. On the one hand, there is broad optimism that AI will be a powerful market DRIVE over the long term; on the other hand, debate never stops over whether and when hundreds of billions of dollars in investment will net profit. This issue is especially sensitive because this spending cycle has turned many big AI investors from powerful cash generators into companies with negative free cash flow. Shea of BNY Wealth Management said: "If there is another situation where free cash flow expectations must be cut, the market will be more worried about the sustainability of AI spending and companies' reliance on new capital, and that worry will also transmit to new cloud providers and other areas of the market. However, if free cash flow expectations can be raised, the market will breathe a sigh of relief, believing that the downward revision process has bottomed and the most difficult phase of the spending cycle has passed." Of course, some seasoned Wall Street professionals still believe that mega-cap tech giants remain one of the more reliable investment choices in the market, especially since this year's declines have already valuations. The "Magnificent Seven" index currently trades at less than 24 times forward earnings, below its 10-year average of about 29 times and below 33 times a year ago. Both Microsoft and Alphabet trade below their respective long-term averages, while Amazon's P/E ratio is below 20 times, near its lowest level since 2008. These lower valuation multiples correspond to fairly strong growth prospects. According to Bloomberg Intelligence data, the "Magnificent Seven" are expected to grow earnings by about 34% this year, above the S&P 500's 29% growth rate. In addition, Wall Street's view of the group's longer-term prospects is becoming more optimistic. The market expects their 2027 profits to grow 27%, a consensus that has been notably revised up in recent weeks; two months ago, analysts expected growth of less than 19%. Tiffany McGee, CEO of Pivotal Advisors, said in a media television interview: "The market has set the bar so high that when these companies meet expectations, the market instead seems disappointed. Investors want them to deliver significantly better-than-expected performance." In addition, the market also expects that hyperscale cloud service providers will continue to push forward with massive AI spending plans regardless of whether they can deliver obvious returns to investorsa scenario that is keeping Wall Street on edge. Kim Forrest, chief investment officer at Bokeh Capital Partners, said: "The 'Big Four' driving new cloud buildout currently seem to have no intention of changing direction at all. They will continue to push forward with spending plans, and that may be making investors nervous." This week, tech giants must prove whether the growth trajectory driven by AI computing infrastructure can outrun financing costs Meta's post-earnings stock reversal revealed the distinction between revenue growth and return on capital. Its future spending commitments rose from about $700 billion to $825 billion, an increase of $125 billion; this is the total commitment for multi-year leases, procurement contracts, and conditional cloud capacity purchases, not one year's capital expenditure, nor does it equal new debt of the same size. For market participants, the specific investment implication behind Meta's latest move is that the company has locked in longer-term spending obligations in advance, so the market needs a clearer path for revenue realization and cash recovery. Equipment investment closely tied to AI computing infrastructure typically creates cash outflows first and then gradually enters the income statement through depreciation, so profit growth and deteriorating free cash flow can absolutely occur at the same time. Amazon provides a concrete example: AWS revenue grew 37% year-over-year to $42.2 billion in the second quarter, but for the trailing 12 months through that quarter, although group operating cash flow grew 33% to $161.4 billion, free cash flow swung from an inflow of $18.2 billion in the year-earlier period to an outflow of $7.6 billion, mainly due to increased AI equipment investment. This means that what this round of earnings needs to prove is that paid demand, computing utilization, and service marginsand even cash flow metricscan show significant improvement signals and gradually cover construction, operating, and financing expenses. For chip and equipment suppliers, capital expenditures still support orders; for cloud providers bearing the investment, shareholder returns depend on how much cash these devices ultimately generate. The market is not without conditions for repair. Institutional estimates show that the forward P/E ratio of the U.S. stock market's "Magnificent Seven" has fallen below 24 times, below the 10-year average of about 29 times, while the 2027 earnings growth expectation has been raised to 27% from less than 19% two months ago. This divergence of "rising earnings expectations and falling valuations" shows that investors are raising their requirements for growth sustainability and capital investment efficiency. If earnings lead to upward revisions in free cash flow expectations, it can simultaneously ease financing concerns and enhance the credibility of profit forecasts; if revenue growth continues to be accompanied by upward revisions in investment commitments and downward revisions in cash flow forecasts, even a lower P/E ratio will be hard-pressed to stabilize the market on its own. This week's Federal Reserve meeting will fully intersect with this earnings testAlphabet and Microsoft will report after the U.S. market close on October 28, and Amazon on October 29 after the close; the Fed will announce its rate decision on the 28th, followed by Warsh's press conference. Under his communication approach of reducing explicit forward guidance, investors need to infer the policy path from changes in inflation, growth, and financial conditions. Tech giants' earnings determine how much future cash return the market is willing to believe in, while Fed policy and the credit market affect the cost of capital for waiting for those returnswhether these two ends can improve simultaneously will be key to repairing the bullish tone of global AI computing-themed and tech stock trading. After record hedge fund selling, strategists at Wall Street financial giant Goldman Sachs suggest "selectively" bottom-fishing AI assets After the recent sharp selloff in AI computing themes, the core judgment of Goldman Sachs' strategist team is that concentrated de-risking has improved the buying conditions for some AI assets, but whether a sustained rally can form still depends on whether earnings growth and cash returns can support investment expansion. Goldman Sachs strategist Louis Miller sees an allocation opportunity formed by lower positioning, falling share prices, and stable earnings expectations for leaders; another strategist team led by Tony Pasquariello worries that funding sources, cost of capital, and investment returns are not yet fully matched. However, both point to a selective AI asset bottom-fishing strategy: look for companies with earnings support still intact and valuations already digested, while retaining hedges, rather than betting on a synchronized reversal across the entire AI sector. This adjustment first manifested as hedge funds actively shrinking core holdings. Goldman Sachs data show that fundamental long/short funds deleveraged on 7 of the past 8 trading days, with beta-adjusted net leverage falling 10.3 percentage points since October to a one-year low; another measure, plain net leverage, fell to 47%, in the 1st percentile of the three-year distribution. Thursday's reduction was entirely driven by long selling, with tech stock reductions the largest single-day amount in about two years, and Goldman Sachs' Hedge Fund VIP basket also posted its worst single-day performance since 2011. All of this means the market is cutting back on what were once its most confident holdings; as exposure declines, the pressure for further concentrated selling may ease, creating repair conditions for companies with stable fundamentals. While capital is withdrawing from core AI holdings, a clear internal rotation is taking place in the market. The Philadelphia Semiconductor Index fell about 6.7% on Thursday, yet 337 S&P 500 constituents rose; the index excluding AI constituents rose 6.5% in October, outperforming the S&P 500 as a whole for the first time this year. Goldman Sachs' AI and AI risk-hedge portfolios and high-beta momentum portfolios have retraced about 55% and 43%, respectively, from their June highs, showing that previously highly concentrated thematic trades and momentum strategies are undergoing severe adjustment. The drawdowns in these specific portfolios reflect the unwinding of crowded trades and cannot be equated with the same percentage decline in the value of the entire AI industry; the investment implication is that the market has begun to reallocate valuation premiums based on earnings quality and capital needs. The Goldman Sachs strategist team remains willing to selectively position, and the most important basis is that earnings expectations for hyperscale cloud service providers (i.e., Hyperscalers including Amazon, Google, and Microsoft) have not been revised down since July, and the data center equipment sector has shown no signs of slowing earnings growth. When share prices fall while expected earnings remain stable, investors pay less for the same earnings. This echoes the phenomenon of the tech "Magnificent Seven" forward P/E falling below 24 times while 2027 earnings growth expectations have been raised to 27%. However, an overall decline in risk exposure does not mean all trades have been fully cleared: Goldman Sachs notes that although momentum exposure has fallen to its lowest level since June 2025, it remains in the 73rd historical percentile, and some popular holdings are still concentrated. The key to stable earnings expectations and intensifying investor concerns coexisting is that earnings per share (EPS) and free cash flow (FCF) reflect different issues. Meta raised the lower end of its annual capital expenditure guidance to $135 billion and increased multi-year future spending commitments to $825 billion, yet its stock turned from gains to losses after earnings, showing that strong revenue alone is no longer enough to support valuation. Equipment purchases consume cash first and then gradually affect profits through depreciation; therefore, even if companies maintain earnings growth, they may need more external funding. In an environment of high interest rates and widening credit spreads, what the market demands to see is that AI paid demand, computing utilization, and service margins continue to improve and can gradually cover equipment investment, operating expenses, and financing costs. This change also explains why Goldman Sachs is focused on behind-the-meter generation and self-built campus power: these areas directly correspond to data center power access and supply bottlenecks, and project progress can relatively specific equipment and service demand. But the existence of engineering solutions for power supply does not mean construction funding and economic returns are already secured. Goldman Sachs is also wary of non-AI stocks rising too fast in the short term and avoids low-quality, cyclical-sensitive assets, reflecting a screening of specific earnings sources. Historically, Goldman Sachs data show that the median return three months after a 10% momentum portfolio drawdown was 16.4%, but after a low in July this year, the portfolio fell a further 15.4%, which also explains why the Goldman Sachs strategist team maintains limited confidence in market direction.