Tech Stocks Shed $1.3 Trillion as AI Spending Paradox Triggers Market Repricing

Technology equities erased over $1.3 trillion in combined market value since January 2026 as investors question whether hyperscaler AI spending will justify stretched valuations, triggering the sharpest sector rotation in three years despite record infrastructure investments.

The selloff has created a stark paradox: Nvidia’s stock is outperforming all megacap peers this year, up 5% as of late February, while Microsoft has lost roughly $613 billion in market value, Amazon shed $343 billion, Apple declined $256 billion, Alphabet dropped $88 billion, and Nvidia itself fell $90 billion from January highs. The Nasdaq Composite dropped 1.55% on February 12, continuing a multi-week decline that has pushed the VIX Volatility index to between 21.01 and 21.53—a 42% spike year-to-date and levels not seen since November 2025.

The market repricing reflects a fundamental shift in investor psychology. Investors are moving from rewarding long-term AI ambitions to demanding near-term earnings visibility after years of speculative enthusiasm, according to Reuters. This transition comes as aggregate annual AI infrastructure commitment from the five largest US cloud companies increased from approximately $380 billion in 2025 to a projected $660-690 billion in 2026—nearly doubling spending in a single year.

The Capex Conundrum

Hyperscalers have committed to unprecedented capital intensity. Alphabet expects 2026 capital expenditures between $175 billion and $185 billion, while Amazon leads with $200 billion in capex for 2026, catching projections off guard as consensus expectations had been closer to $147 billion. Microsoft and Meta round out the cohort with massive investments of their own.

The scale is historically unprecedented. Tech CapEx as a percentage of GDP nearly matched the combined scale of the largest capital projects of the 20th century, reaching approximately 1.9% of GDP in 2025—exceeding nationwide broadband development at the start of the century, which represented 1.2% of GDP, according to analysis by IEEE ComSoc.

But the spending is consuming cash flow at alarming rates. Hyperscaler capex in 2026 will consume nearly 100% of operating cash flows, compared to a 10-year average of 40%, according to UBS. This has forced companies to tap debt Markets: The Big Five hyperscalers raised over $121 billion in new debt during 2025, with projections suggesting the sector may need more than $400 billion in borrowing in 2026, more than double 2025’s $165 billion.

Nvidia’s Exception Proves the Rule

Nvidia reported fiscal 2026 fourth-quarter earnings of $1.62 per share on revenue of $68.1 billion and guided to fiscal 2027 first-quarter revenue of $78.0 billion, topping Wall Street expectations across all metrics. Revenue in the data center business climbed 75% from a year earlier to $62.3 billion. Yet even this performance couldn’t sustain broader market enthusiasm—the stock initially surged 3.5% after hours before giving back gains the following session.

The bifurcation is instructive. Investors have rotated away from AI infrastructure companies where operating earnings growth is under pressure and where capex is being funded via debt, while rewarding companies demonstrating a clear link between capex and revenues, according to Goldman Sachs Research. Since June 2025, the average stock price correlation across large public AI hyperscalers has collapsed from 80% to just 20%.

Revenue Mismatch Fuels Skepticism

The fundamental issue troubling investors: AI revenue doesn’t match infrastructure investment. OpenAI’s $20 billion ARR represents roughly 3% of projected 2026 hyperscaler capex, while Anthropic’s $9 billion run rate occupies a similar position. The entire cohort of pure-play AI vendors likely accounts for less than $35 billion in projected combined 2026 revenue against nearly $700 billion in capex.

Spending on AI infrastructure now forms a significant part of US GDP growth, and corporate capital expenditure would be negative without it, according to Pantheon Macroeconomics. Overall capex rose 2.6% in Q4 2025, but within that, intellectual property and software spending linked to AI was up 7.4%, while computer and communications equipment surged 61%. All other segments declined, with investment in other equipment plunging 17%.

Analyst Split: Peak or Plateau?

Wall Street remains divided on whether AI capex has peaked. Goldman Sachs analysts predicted that the runaway quarterly growth rate of AI infrastructure capex is likely to decelerate in late 2026, according to Investing.com. As a result, revenue growth and valuations of stocks viewed as providers of AI infrastructure appear vulnerable to slowdown.

Yet Nvidia CEO Jensen Huang’s message to investors was clear: Big Tech’s massive spending on AI technology is not anywhere near finished, stating “this new way of doing computing is not going to go back” and businesses will “continue to expand from here”. This sets up a binary outcome: either monetization accelerates dramatically, or the market reprices further.

Hyperscalers spent $106 billion in capex in Q3 2025, representing 75% year-over-year growth, though analysts expect this to slow to 49% in Q4 and 25% by end of 2026. However, consensus estimates have consistently underestimated capex for two years running—at the start of both 2024 and 2025, estimates implied roughly 20% growth, but actual spending exceeded 50%.

Comparison to Prior Cycles

The current buildout invites comparisons to the dot-com bubble and mobile/cloud transitions. During the dot-com era, telecom companies overbuilt fiber capacity that took years to absorb. The 5G telecom buildout consumed about 70% of operating cash flow at its peak—similar to where AI infrastructure spending stands today, according to Real Investment Advice.

The cloud infrastructure cycle offers a more optimistic parallel—initial heavy investment eventually generated sustained returns as enterprises migrated workloads. But that transition took nearly a decade to fully mature, and competition consolidated to three major players.

What to Watch

Upcoming hyperscaler earnings through March will provide critical signals on whether capex-to-revenue ratios are stabilizing or deteriorating further. Microsoft reported $625 billion in RPO backlog growing at 110%, Oracle showed $523 billion up 438%, while AWS and Google Cloud each reported $240 billion backlogs—suggesting demand remains robust.

Three indicators will determine whether February 2026 marks peak fear or the start of sustained repricing: first, whether Q1 2026 cloud growth rates stabilize above 20% across providers; second, whether pure-play AI vendors can demonstrate a path to profitability within 18 months; and third, whether AI productivity gains begin appearing in non-tech corporate earnings reports.

The VIX’s decline from 21+ levels to around 19 by late February suggests some fear has dissipated post-Nvidia earnings. But with 49% of public tech companies down at least 25% from 52-week highs and capital intensity at decade highs, the market is pricing in a show-me moment. The $700 billion question: can AI infrastructure investments generate returns fast enough to justify valuations before the next cycle turns?