On Thursday, July 23, 2026 (Eastern Time), all three major U.S. stock indexes closed lower. At the close, the Dow Jones Industrial Average dropped 506.93 points to 51,711.65, down 0.97%. The S&P 500 fell 90.66 points to 7,408.30, a decline of 1.21%. The Nasdaq Composite plunged 553.21 points to 25,137.69, down 2.15%. The "Magnificent Seven" tech index slid about 4.8% for the day, with the combined market capitalization of these seven companies evaporating by roughly $797 billion—their worst single-day performance since the tariff-driven sell-off in April 2025.
The immediate trigger for this sell-off was the earnings reports from two tech giants. Alphabet, Google’s parent company, saw its Q2 capital expenditures double year-over-year to $44.9 billion and raised its full-year 2026 capex guidance to between $195 billion and $205 billion. Tesla’s Q2 revenue grew 26% year-over-year to $28.24 billion, but adjusted earnings per share came in at just $0.33—well below the market expectation of $0.50. Its operating margin fell sharply to 1.4%. Together, these reports reinforced a deep-seated market concern that has persisted for weeks: Big Tech is pouring hundreds of billions into AI infrastructure at an unprecedented pace, but the returns may not yet be sufficient to justify such aggressive spending.
The AI investment boom has now lasted three years, delivering some of the strongest returns for U.S. stocks in decades, but also leading to an unprecedented concentration in the market. Wall Street remains divided on a central question: Is this an unsustainable bubble, or the dawn of a technological revolution reshaping the global industrial landscape?
Bullish Perspective: Real Compute Shortages, AI Infrastructure Still in Early Stages
The bullish case is built on fundamental supply and demand. Juan Correa, Head of Portfolio Construction at BCA Research, and Noah Weisberger, Head of Equity Research, note that the compute market is still supply-constrained, not oversupplied, and that the market is actually underestimating—not overestimating—the returns on capital expenditures. The key supporting data comes from cloud business backlogs: Signed but unrecognized revenue orders at hyperscale cloud providers have grown by approximately $750 billion over the past two quarters, indicating that industry demand continues to outpace supply.
On the demand side, AI adoption is accelerating. In Q2, Alphabet’s Google Cloud revenue jumped 82% year-over-year to $24.8 billion, beating analyst expectations of $22.3 billion. Gemini’s monthly active users reached 950 million, with daily active users tripling over the past year. Gemini models now process 22 billion API tokens per minute, and nearly 90% of Fortune 100 companies in the U.S. have adopted Gemini Enterprise. Microsoft’s annualized AI revenue run rate has surpassed $37 billion, up 123% year-over-year.
On the supply side, compute shortages remain significant. Morgan Stanley analysts expect memory prices to rise at least 25% in Q3 2026 compared to Q2, with supply constraints likely to persist through 2028. In 2026, the four major overseas cloud providers are projected to spend a combined $725 billion on capital expenditures, while advanced packaging, high-bandwidth memory, and high-end chips still face supply-demand gaps of 20% to 45%. Ben Snider, Chief U.S. Equity Strategist at Goldman Sachs Research, recently pushed back against AI bubble concerns, arguing that the risk of a sharp pullback in hyperscale data center spending is low and that AI infrastructure investment will continue to grow.
From a valuation standpoint, bulls argue that we are not at typical bubble levels. As of July 23, NVIDIA’s price-to-earnings (TTM) ratio stood at 31.49. By comparison, the Nasdaq’s peak P/E ratio during the dot-com bubble exceeded 150. Current valuations for leading AI hardware companies are still well below those historical highs. Hyperscale cloud providers are trading at their lowest valuations in nearly a decade, while their returns continue to improve.
Bearish Perspective: Massive Capex Erodes Cash Flow, Valuations Reflect Best-Case Scenarios
The bearish camp also grounds its argument in solid financial data. Peter Berezin, Chief Economist at BCA Research, and Arthur Budaghyan, Chief Strategist for Core Macro and Emerging Markets, believe U.S. stocks are currently in a profit bubble—headline profit figures look strong, but profit margins are much less sustainable than they appear.
Valuation data is cause for concern. If S&P 500 profit margins revert to 2019 levels, the index’s current forward P/E would be about 27—already above the historical peak of 26.5 reached during the dot-com bubble in March 2000. Even excluding tech and financial stocks, the rest of the index trades at a rolling P/E of 26, while profit growth over the past two to three years has averaged only about 3%.
The most pressing concern for bears is that capital expenditures are eroding free cash flow. Alphabet’s free cash flow in Q2 dropped to negative $5.9 billion—the first negative figure in decades. The company not only raised its full-year capex guidance to as high as $205 billion, but also signaled that spending will continue to surge in 2027. To fund this round of AI infrastructure buildout, Alphabet has issued $49.6 billion in new financing, sold $20.3 billion in senior unsecured bonds, and is preparing a $40 billion ATM stock offering. Microsoft’s Q3 FY2026 capex was $31.9 billion, with guidance for Q4 to exceed $40 billion, and total 2026 calendar year capex expected to reach about $190 billion. Microsoft’s free cash flow for the same period was $15.8 billion, down 22% year-over-year, mainly due to the $31.9 billion invested in AI infrastructure. Amazon’s 2026 capex is projected at about $200 billion, with Q1 capex already at $44.2 billion—free cash flow plummeted 95% to $1.2 billion as a result. Meta raised its 2026 capex guidance to between $125 billion and $145 billion, with the midpoint up about 8%. Morgan Stanley projects that by 2026, hyperscale cloud providers will account for about 40% of total capital expenditures in the Russell 1000—double the share in 2024.
Bears further point out that while massive capex boosts chip suppliers’ revenues and profits on paper, hyperscale cloud providers account for these chip purchases as capital expenditures rather than operating costs. This means that reported profits rise without a corresponding increase in cash flow. JPMorgan recently issued a warning that there is significant divergence within the U.S. AI sector, reminiscent of the late stages of the dot-com bubble, which could trigger a sharp correction. Strategist Jason Hunter noted that today’s market divergence closely mirrors the late 1990s. JPMorgan Asset Management Chairman Cembalest put it this way: "The carriage (infrastructure) is running ahead of the locomotive (front-end applications)," a classic sign of a market top.
Historical Context: Where Are We in the AI Cycle?
To assess where AI stands today, it’s useful to compare its position within the broader arc of technological cycles in history.
Similarities and Differences with the Dot-Com Bubble (1995–2000). The parallels are clear: narrative-driven markets, widespread belief that new technology will change the world, capital concentrated in a handful of growth stocks, and soaring valuations for industry leaders. Morgan Stanley acknowledges that today’s AI rally does share many traits with the late stages of the dot-com bubble.
But the differences are just as significant. Since the launch of ChatGPT, the Nasdaq 100 has gained over 140%. During the 1995–2000 dot-com era, the Nasdaq surged nearly 600%. The current rally’s slope is far less steep. More importantly, the AI sector now enjoys real revenue support—Alphabet’s Google Cloud posted $24.8 billion in quarterly revenue, and Microsoft’s annualized AI revenue run rate exceeds $37 billion. Most ".com" companies during the bubble never had such tangible business results.
Reference to the Cloud Computing Cycle (2008–2015). Analysts believe AI adoption will follow a curve more like cloud computing’s 5–10 year rapid trajectory, rather than the 20-year PC adoption cycle, and will be amplified by the scale of potential productivity gains. In the cloud era, capital expenditures by giants like Amazon, Microsoft, and Google grew from $2 billion in 2004 to an estimated $159 billion in 2026—an 80-fold increase. The investment intensity of AI infrastructure far exceeds that of early cloud computing: In 2026, the world’s top five cloud providers are expected to spend over $800 billion in total capex, potentially rising to $1 trillion–$1.2 trillion in 2027. This level of capital intensity is both a testament to AI’s potential and a source of its fragility.
Conclusion
The market sell-off on July 23, 2026, was essentially a collective stress test of capital markets’ expectations for AI investment returns. The bull-bear divide is no longer about whether AI is transformative—that’s now a given—but whether the scale and timing of capital expenditures are aligned with the eventual returns.
Bulls can point to $750 billion in backlog growth, 82% cloud revenue increases, and 20–45% chip supply-demand gaps as evidence of robust demand. Bears, meanwhile, cite a 27x cyclically adjusted P/E, negative or sharply declining free cash flow at major tech firms, and over $800 billion in planned 2026 capex as supply-side and financial red flags.
History offers two contrasting reference points: The dot-com bubble warns of the dangers of valuations decoupling from fundamentals, while the cloud computing era shows that massive early infrastructure investments can translate into sustained returns over the following decade. For investors, the most crucial metric right now may not be any single company’s quarterly earnings, but rather when AI infrastructure capex begins to drive verifiable improvements in free cash flow. That turning point is when the real bull-bear showdown will be decided.
FAQ
Q1: Are current AI stock valuations already in bubble territory?
By some measures, they are at historical highs—the S&P 500’s cyclically adjusted P/E is around 27, surpassing the 26.5 peak during the dot-com bubble in 2000. However, leading AI companies generally have real revenue and profit to back up their valuations, unlike most companies during the dot-com era that had only concepts and no actual sales. High valuations don’t guarantee a bubble will burst; the key is whether earnings growth can continue to support those valuations.
Q2: What are the main risks of Big Tech’s massive AI capital expenditures?
The core risk is a long investment payback period. Alphabet’s 2026 capex guidance has been raised to $195–205 billion, and its free cash flow has turned negative. Microsoft, Amazon, and Meta are all facing significant cash flow compression due to heavy capex. If commercialization of AI applications lags expectations, these massive outlays could erode shareholder returns for years.
Q3: When will AI infrastructure investments deliver verifiable returns?
We’re still in the early stages. Alphabet CEO Sundar Pichai has stated that "AI returns are still in their early phase." Microsoft’s annualized AI revenue run rate has reached $37 billion, and Google Cloud revenue is up 82% year-over-year. The market is closely watching for a sustained improvement in free cash flow—a key signal that AI investment is moving from the "spending phase" to the "returns phase."
Q4: How is the correlation between crypto assets and AI stocks changing?
Over the past month, Bitcoin has largely tracked the trading patterns of AI stocks. However, while tech stocks took a beating on July 23, Bitcoin hovered around $65,000 during Asian trading on July 24, with an intraday drop of less than 1%. The market is watching to see if this signals a new phase where Bitcoin and AI stocks begin to decouple.
Q5: Is the current AI rally more like the dot-com bubble or the cloud computing cycle?
Both historical comparisons are instructive. The AI rally resembles the dot-com bubble in its narrative-driven momentum and market concentration, but it also mirrors the cloud computing cycle in its infrastructure-heavy investment model and relatively rapid adoption curve. Today’s AI sector has real revenue foundations absent during the dot-com era, but its capex intensity far exceeds the early days of cloud computing. In short, AI may be in a unique phase that combines the "narrative intensity of the dot-com bubble" with the "investment model of the cloud era."
