Why did Google lose $225 billion in a single day? AI talent drain and $760B in capital expenditure trigger a reshaping of big tech valuations

On June 22, 2026, Alphabet (GOOGL.O) shares saw their largest single-day drop since May 2025. Google-A (GOOGL) closed down 4.99% to $349.68, while Google-C (GOOG) closed down 5.08% to $348.78. During the session, the decline briefly widened to 7.2%, the biggest intraday drop since February. Based on the intraday low, the market value had at one point evaporated by about $320 billion. Even using closing prices, the daily market-cap loss still reached roughly $225 billion.

This drop was not an isolated incident. On the same day, Bloomberg’s “Magnificent Seven” index fell as much as 2.2%. Amazon fell 4.75% to $232.79, Microsoft dropped 3.18% to $367.34, and Meta Platforms slid 2.32% to $563.85. Large-cap tech stocks were under pressure across the board.

Market interpretations centered on two aspects: a continuous outflow of core AI talent, and structural anxiety over AI infrastructure capital expenditures. These two narratives resonated on the same day, pushing Google to the intersection of public-opinion pressure and valuation pressure.

Talent Drain: The Significance of a Nobel Laureate Exiting

Over the past week, Google DeepMind has lost two top AI researchers one after another. On June 18, Noam Shazeer, Google’s Vice President of Engineering and co-lead of the Gemini model, announced his departure and said he would join OpenAI. Shazeer left Google in 2021 to found Character.AI, returned to Google in 2024 via a deal of about $2.7 billion, and then left again less than two years after his return.

Just two days later, John Jumper, Vice President of Google DeepMind, announced his departure and said he would join Anthropic. Jumper, together with DeepMind CEO Demis Hassabis, won the 2024 Nobel Prize in Chemistry. AlphaFold—work he helped lead—has predicted more than 200 million protein structures. Jumper spent as long as nine years at Google and was a core member of the company’s AI coding development team.

The back-to-back departures of these two top talents triggered market worries that Google is “losing the AI talent war.” Gil Luria, Managing Director at DA Davidson, noted that last year Google had the most advanced models and won market recognition as an AI winner, but has since gradually fallen behind; these departures could mean it might fall even further. Citizens analyst Andrew Boone traced a longer-term concern—years ago, one of the market’s main worries about Google was the continued flow of AI talent to new competitors, which weakens its ability to catch up technologically.

Capital Expenditure Anxiety: When the Gap Between Investment and Returns Is Amplified

Beyond talent loss, more structural pressure comes from the scale of capital spending on AI infrastructure. In 2026, the combined capital expenditures of the five largest hyperscale cloud providers—Alphabet, Amazon, Meta, Microsoft, and Oracle—are expected to reach about $760 billion. Goldman Sachs estimates this figure at roughly $770 billion, or about 100% of these companies’ operating cash flow.

Alphabet’s own capital expenditure guidance for 2026 is $175 billion to $185 billion. Since October 2025, the company has raised about $141 billion through debt and equity financing to expand AI infrastructure.

However, the expansion in spending has not been matched by improvements in cash flow on a comparable basis. Alphabet’s free cash flow in Q1 2026 fell 47% year over year to $10.12 billion. The combined free cash flow of the five hyperscale cloud providers is expected to plunge 91% from current levels to about $16 billion. Amazon and Oracle are expected to post negative free cash flow in 2026.

This set of data forms a clear narrative: building AI infrastructure is consuming cash at an unprecedented pace, but the pace of monetization has not yet caught up. Goldman Sachs warns that depreciation and amortization expense for hyperscale cloud providers will rise as a share of revenue from 7% in 2022 to 12% in 2027, and that the return on shareholders’ equity is expected to average a decline of 7 percentage points next year.

Commercialization Anxiety: Nadella’s “Commoditization” Remark

In an interview with The Wall Street Journal on June 22, Microsoft CEO Satya Nadella said the artificial intelligence market is moving toward commoditization, and the industry should not overly rely on only a few “AI giants.” This statement directly challenges the market’s “winner-takes-all” pricing logic for AI leaders such as Google.

If AI models become cheaper and more substitutable, can Google’s vertically integrated AI system form a long-term moat—or will it simply translate into pressure on profit margins? This is the question investors repeatedly asked during the June 22 trading session.

Restructuring the AI Compute Narrative in Large-Cap Portfolio Weights: From “Winner-Takes-All” to Benefiting Equipment Suppliers

Google’s plunge is not just a single-stock event. It reflects a structural restructuring in the AI compute narrative—markets are reallocating winners and losers in the AI investment wave.

The logic of divergence: device makers vs. cloud service providers

In a Goldman Sachs report released in June, the bank explicitly pointed out that clear divergence is emerging in the AI infrastructure buildout wave. Semiconductor manufacturers are among the biggest beneficiaries of the AI spending surge, with industry net profit margins nearing 50%. Companies such as Nvidia, Micron Technology, and Broadcom continue to benefit from strong demand alongside a supply-constrained landscape.

Meanwhile, hyperscale cloud providers bear the enormous costs of the AI infrastructure race. This spending is reshaping their financial structures—asset turnover declines, depreciation expenses rise, and debt and equity financing increase.

The trading data from June 22 confirmed this divergence. Micron rose 6.82% to $1,211.38, and Intel rose 5.19% to $140.94. Gil Luria of DA Davidson summarized it this way: investors are selling the companies that spend on AI compute, while buying the companies that get paid for it.

Implicit adjustment of large-cap portfolio weights

The divergence in the AI compute narrative is reshaping the relative weights of large-cap stocks. The seven biggest tech giants currently collectively deliver a 44% return on shareholders’ equity, up 9 percentage points from three years ago. But the sustainability of this figure is being questioned.

The core issue lies in timing effects in accounting treatment. Equipment manufacturers recognize revenue immediately when sales occur, while cloud providers allocate data center construction costs as depreciation over future years. This means that part of the current earnings growth—S&P 500 constituent companies have delivered more than 20% earnings growth for two consecutive quarters—is built on an accounting window.

Morgan Stanley accounting analyst Todd Castagno described it as “a golden window where everyone looks good.” But Visible Alpha data shows that among the five hyperscale cloud providers, there is an approximately $5.49 billion gap between 2026 capital expenditures (about $760 billion) and depreciation and amortization (about $211 billion)—costs have already been incurred, but they have not yet been reflected in the income statement.

As this depreciation gradually comes due in financial reporting, the quality of large-cap earnings and the valuation logic will need to be re-examined. The S&P 500’s current forward price-to-earnings ratio of about 22x is above the historical average, and the divergence in the AI compute narrative adds new uncertainty to the sustainability of this valuation.

When Traditional Stocks Meet Crypto Infrastructure: How Gate Bridges the Gap

Against the backdrop of the restructuring in large-cap weights driven by the AI compute narrative, the boundary between traditional stocks and digital assets is becoming increasingly blurred. As a pioneer of this convergence trend, Gate.io offers investors the possibility to manage cryptocurrencies and stock assets within a unified ecosystem.

Real Stock Trading: USDT Enters Directly

On June 1, 2026, Gate officially launched real stock trading services, becoming one of the first exchanges in the industry to directly connect to the US stock market within a crypto platform. Users do not need to exchange currencies, do cross-border remittances, or open an additional brokerage account. They only need to use the USDT liquidity in their Gate account to buy real stocks listed on major US exchanges such as the New York Stock Exchange and Nasdaq with a single click.

The core innovation of this model lies in combining blockchain-native settlement with exposure to traditional stock holdings. Users do not need to deposit fiat or perform currency exchange; instead, they operate via crypto balances while also gaining exposure to underlying stock price movements.

Trading 7×24 Hours: Breaking Through Traditional Trading Session Limits

On June 22, 2026, Gate announced that stock trading has been fully supported for 7×24-hour trading around the clock. Building on the existing pre-market, regular-hours, and after-hours trading, it added overnight sessions and weekend trading, covering US stocks, Hong Kong stocks, and Korean stocks.

This means investors can adjust positions during periods when traditional markets are closed—including the Asian trading session after Google’s plunge. For investors who want to react immediately after major news events (such as announcements of AI talent departures), this feature has clear practical value.

Fragmented Investing and Low Barriers

Thanks to the divisibility of blockchain technology, Gate supports fractional-share trading with a minimum of 0.01 share. For example, taking Google as an example: even if the per-share price is above $340, investors can still participate with far less than the funds needed for a full share. The Gate stocks section has already added nearly 100 trading pairs, covering multiple segments such as tech giants, leading companies in aerospace and defense, consumer goods leaders, and core ETFs.

Dual-Track System: Real Stocks and Tokenized Stocks

Gate offers two trading routes: real stocks and tokenized stocks (xStocks and Ondo Stocks). Tokenized stocks are pegged 1:1 to real stock prices. Users can buy and sell nonstop 7×24 hours like trading crypto assets, with no restrictions from traditional stock market closures and holidays. Real stock trading provides direct ownership exposure, suitable for investors who prefer traditional asset structures.

Both modes share the same account and asset system. Users only need to complete basic KYC to begin trading. Holdings of $2,000 can be upgraded to VIP status, enjoying the lowest 0.023% stock trading fee rate.

Capturing Trading Opportunities in the AI Narrative Restructuring

As the AI compute narrative spreads from “model winners” to “infrastructure suppliers,” Gate’s multi-asset trading capabilities give investors flexible allocation tools. Whether directly trading the stocks of hyperscale cloud providers such as Google, Microsoft, and Amazon, or participating in price movements of equipment suppliers such as Nvidia and Micron Technology through tokenized stocks, investors can complete cross-asset-class allocation adjustments within a single unified interface.

Conclusion

Google’s plunge on June 22, 2026 marks a concentrated release of three narratives: AI talent outflow, capital expenditure anxiety, and commercialization concerns. A 7% intraday decline, $320 billion in market-cap evaporation, and the consecutive departures of two Nobel-level AI researchers—all these figures point to a deeper trend: the AI compute narrative is shifting away from the simple “winner-takes-all” logic toward a more complex divergence between equipment suppliers and cloud service providers.

The relative weights of large-cap stocks are being reassessed. When the combined capital expenditures of the five hyperscale cloud providers reach $760 billion, yet free cash flow could plunge 91%, the market has no choice but to re-examine the match between earnings quality and valuation logic.

Amid this structural change, the integration of traditional stocks and digital assets provides investors with new trading dimensions. Gate.io’s dual-track system of real stocks and tokenized stocks, along with 7×24-hour trading capability, enables investors to maintain higher operational flexibility and asset allocation efficiency as the AI narrative is reshaped.

The story of AI compute is far from over—but the way it is told is changing.