Why Are AI Stocks Falling Today? Nvidia, AMD, Micron, and Broadcom Lead Chip Stock Declines

On September 14, AI stocks came under pressure once again, but the decline was clearly concentrated in one part of the market: semiconductors and AI infrastructure.

Nvidia fell 3.4%, Micron Technology dropped over 5%, and both AMD and Broadcom declined more than 4%. The Philadelphia Semiconductor Index fell 5.9%, far exceeding the 0.56% decline in the Nasdaq Composite. The market is refocusing on the pace of frontier AI development and safety risks, while higher U.S. Treasury yields are adding further pressure on highly valued tech stocks. Reuters reported that chip stocks led Wall Street's decline

The market reaction does not necessarily mean investors believe the AI boom is about to end. The more specific question is: if the pace of frontier AI development slows, what happens to demand for GPUs, HBM, networking equipment, and data center capacity?

Key Takeaways

• Nvidia fell 3.4%, while Micron dropped over 5%.
• Both AMD and Broadcom declined more than 4%.
• The Philadelphia Semiconductor Index fell 5.9%, narrowing its 2026 year-to-date gain to 57%.
• The Nasdaq fell only 0.56%, indicating that selling pressure was concentrated in semiconductor and AI infrastructure stocks.
• Investors are reassessing whether frontier AI safety concerns could ultimately affect the pace of growth in AI infrastructure spending.

Why did AI chip stocks fall today?

The direct catalyst came from the market repricing the pace of frontier AI development.

Recent AI safety incidents have made this discussion more concrete. OpenAI said earlier this month that GPT-6 Astra became its first widely deployed model to reach the company's Critical cybersecurity capability threshold. OpenAI also said it delayed some development and release processes for Astra while strengthening safeguards against cyber abuse and unauthorized model behavior. OpenAI's Path to Astra safety assessment

Anthropic has also published similar progress on frontier model capabilities. Its latest research found that some AI systems are already capable of performing tactical intelligence and conventional weapons-related tasks that previously typically required scarce, highly trained human experts. Anthropic said that as model capabilities continue to advance, these results highlight the need for stronger safety mechanisms. Anthropic's latest frontier capability assessment

These concerns became further catalysts for the stock market after leaders at Anthropic, OpenAI, and xAI warned about the risks of rapid AI development and supported slowing down some aspects of development. The semiconductor sector reacted most sharply because many AI chip companies' valuations depend heavily on market expectations that compute demand will continue to expand at a rapid pace. Reuters coverage of the September 14 AI chip stock selloff

Why are Nvidia, AMD, Micron, and Broadcom more vulnerable?

If the AI supply chain is broken down into different layers, this round of selling becomes easier to understand.

Nvidia and AMD provide the accelerators needed for AI computing. Broadcom has significant business exposure to AI networking and custom chip infrastructure. Micron supplies memory products, including high-bandwidth memory used alongside advanced AI accelerators. To further understand the relationships between these companies, you can refer to MEXC's AI semiconductor supply chain guide, which explains how chip designers, foundries, equipment makers, and memory suppliers together form the AI infrastructure cycle.

Micron's exposure differs slightly from that of Nvidia or AMD, because memory demand depends not only on how many AI accelerators are deployed, but also on how much and what type of memory each hardware generation requires. MEXC's guide to the differences between DRAM, NAND, and HBM explains why HBM has become one of the memory products most closely correlated with AI accelerator demand.

Therefore, these stocks can fall even without the market believing that AI demand will disappear. Even a small change in expectations is enough to have an impact. If investors begin to expect slower growth in data center construction, GPU deployment, or HBM demand, companies upstream in the infrastructure chain could experience larger valuation adjustments than cloud computing and software companies that actually use these technologies.

Does this selloff mean AI infrastructure demand is slowing?

There is currently little evidence that large-scale AI infrastructure projects are being canceled on a massive scale.

At the end of August, AWS and Nvidia announced plans to deploy an additional 2 million Nvidia GPUs on AWS infrastructure during 2027 to 2028, covering Blackwell Ultra, Rubin, and Rubin Ultra systems. The two companies said customer demand has already exceeded previous expectations as agentic AI, scientific computing, enterprise automation, and physical AI workloads expand. AWS and Nvidia's 2 million GPU expansion plan

Nvidia has also partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish financing platforms, with plans to mobilize more than $500 billion in third-party capital for AI infrastructure over the long term. This also indicates that the industry is still preparing for large-scale expansion of computing capacity, rather than facing an immediate collapse in demand. Nvidia's AI infrastructure financing plan

Therefore, the more important question is not whether AI spending will disappear, but whether its growth rate is changing.

This distinction is very important for valuations. Semiconductor stocks had previously benefited from market expectations that demand for GPUs, HBM, and networking equipment would continue to accelerate. Even if the absolute scale of AI spending remains at historic highs, these companies could still be repriced if the market begins to believe that the same infrastructure buildout will proceed at a slower pace.

Why did chip stocks fall far more than the broader market?

September 14 also highlighted an important divergence within the tech sector.

The Nasdaq fell 0.56%, while the Philadelphia Semiconductor Index dropped 5.9%. Meanwhile, several software stocks such as ServiceNow, Adobe, and Workday actually rose. This suggests investors are not selling all AI-related companies, but rather reassessing which parts of the chain are most vulnerable if AI infrastructure spending becomes more restrained. Reuters September 14 market recap

The logic is not complicated. When cloud computing companies aggressively increase compute capacity, AI infrastructure suppliers benefit directly; but the cloud computing and software companies buying this equipment must also bear the costs. If the pace of frontier AI development slows, hyperscalers may gain more time to monetize existing capacity before entering the next round of spending.

As a result, the risks facing Nvidia, AMD, Micron, and Broadcom are not the same as those facing companies that monetize AI through cloud services, advertising, enterprise software, or consumer applications.

Is the AI boom over?

There is currently no evidence that the AI infrastructure cycle has ended.

Large GPU deployments are still being planned, capital continues to flow into data center infrastructure, and major tech companies are still developing new AI products. What has truly changed is that the market is no longer willing to assume that infrastructure demand can accelerate indefinitely.

This continues a broader shift that has already emerged in AI stocks recently. MEXC previously discussed this issue in Is the AI Tech Sector Entering a Bear Market? The core remains similar: investors are shifting from simply rewarding "AI concept exposure" to demanding clearer evidence of revenue growth, usage rates, and return on capital.

Therefore, the September 14 selloff looks more like the market repricing the pace of the AI infrastructure cycle, rather than rejecting artificial intelligence itself.

What could drive AI stocks next?

The next real validation will come from corporate guidance, not a single trading day.

Investors will be watching whether major cloud service providers change their capital expenditure plans, whether accelerator orders for Nvidia and AMD show changes, whether Micron continues to report strong HBM demand, and whether networking equipment and data center suppliers can maintain their growth expectations.

The divergence between hardware and software is also worth continuing to watch. If semiconductor stocks remain under pressure while cloud computing and software companies relatively outperform, the market may be shifting from a broad "buy all AI" trade to a more selective judgment: who can earn returns from AI spending, and who bears the cost of building the infrastructure.

This would represent an important change in the AI trade—not from growth to collapse, but from enthusiasm to validation.

FAQ

Why did Nvidia stock fall today?

Nvidia fell 3.4% on September 14 as semiconductor stocks were sold off that day amid renewed market focus on the pace of frontier AI development. Investors are assessing whether a slower AI development cycle could ultimately reduce the growth rate of new AI infrastructure spending. Reuters reported on Nvidia's September 14 decline

Why did Micron stock fall today?

Micron dropped over 5%. The company is highly exposed to the AI infrastructure cycle through advanced memory products such as HBM, so changes in expectations for GPU deployment and AI server demand also affect the market's assessment of memory growth.

Why did semiconductor stocks fall?

The Philadelphia Semiconductor Index fell 5.9% as the market grew concerned about the risks of rapidly developing frontier AI and reassessed how these concerns could affect future demand for GPUs, memory, networking equipment, and other AI infrastructure. Higher U.S. Treasury yields also added pressure, but semiconductor stocks fell significantly more than the broader market. Reuters full market coverage of the chip stock selloff

Is the AI boom over?

There is currently no evidence that AI infrastructure demand has collapsed. Major companies are still planning large-scale AI deployments, including AWS and Nvidia's plan to add 2 million GPUs. The current debate is more about the future growth rate of spending, rather than AI demand itself disappearing.