Nasdaq sinks as AI infrastructure drops; Nvidia falls 16% in AI rout

U.S. stocks mostly fell today, and the Nasdaq took the biggest hit in a broad AI rout sparked by China’s DeepSeek. The spotlight quickly turned to the AI infrastructure names, where declines were steep and often landed in double digits, with Nvidia down 16%. That is the kind of percentage move that does not stay contained to one ticker, because Nvidia is not just another semiconductor stock. It is a core plumbing supplier for the AI buildout, meaning when it drops that hard, investors recalibrate the entire “who benefits from AI spending” map.

The timing matters because the story is not “AI is down.” It is “AI infrastructure is down, and the pressure looks synchronized across the stack.” When Nvidia falls 16% in a day while much of the market is only “mostly lower,” the implication is that traders are targeting a specific risk: that near-term AI demand, pricing power, or growth assumptions may be less certain than the market priced in. In other words, this selloff is about expectations tightening, not about a gradual digestion of gains.

To understand why infrastructure names got torched first, you have to know how the market thinks about AI spend. AI model makers need compute, data pipelines, networking, and a whole industrial ecosystem to turn research into revenue. Public markets tend to treat the biggest, most widely deployed vendors in that ecosystem as the transmission mechanism for AI capex. So if investors believe the buildout will slow, shift, or change shape, they often express it through the infrastructure leaders first. Nvidia down 16% is the clearest expression of that impulse.

Then comes the catalyst that shaped sentiment: the rout was sparked by China’s DeepSeek. Even though today’s move is about U.S.-listed stocks, the market reaction is global because AI progress is too. When a China-linked AI development becomes a headline, it can trigger fast reassessments of competitive advantage. Those reassessments might be about the performance per dollar, the efficiency of deployments, or the pace at which capabilities improve. The exact mechanism may be fuzzy in any one press cycle, but the trading behavior is not: when a headline suggests the competition is getting smarter or cheaper, the market starts asking whether everyone else’s current cost structure and roadmap remain the right ones.

Regulatory and policy context adds another layer to how this plays out in capital markets. AI is not just a tech category anymore, it is a policy category. Export controls, national security framing, procurement priorities, and cross-border scrutiny can all affect how quickly compute hardware can be bought, shipped, and deployed. That means executives in AI infrastructure are constantly planning for uncertainty around supply, compliance, and market access. A headline-driven selloff can become a volatility amplifier, because it invites both narrative and risk re-pricing, especially for firms tied to global demand and procurement cycles.

The second-order effect for boards and CFOs is that expectations can shift faster than operating realities. Markets can move ahead of results by days or weeks, particularly when the move is “broad” and touches multiple AI infrastructure names with many double-digit declines. That can impact not only the stock price, but also the cost of capital, employee retention dynamics (through equity), and the willingness of customers to commit to large upgrade cycles. If customers perceive that the market is questioning the economics of AI, buying committees can pause, renegotiate, or stretch timelines.

For peers with similar exposure to AI infrastructure demand, the strategic stakes are direct. If the selloff reflects a belief that AI compute demand growth will be less smooth than expected, leaders need to focus on proof points that underwrite long-term spending: measurable performance, deployment readiness, and customer retention. If it reflects a “competition gets cheaper” storyline linked to developments in China, leaders may need to articulate what differentiates their stack beyond raw benchmarks: software ecosystem, reliability, supply continuity, and total cost of ownership.

Right now, the only hard thing the market is agreeing on is the direction: U.S. stocks were mostly lower, the Nasdaq led the declines, and AI infrastructure was hit hardest, with Nvidia down 16%. That kind of synchronized drop is a reminder that in AI, capital markets can swing from exuberance to anxiety quickly, and the companies most wired into AI infrastructure are often the ones that feel it first.