AI stocks are under pressure as top executives call for slower development, but is this really the end of the AI rally?

AI stocks came under pressure after several leading AI executives called for a slower pace of development, raising fresh questions about how long the current investment boom can continue. The market reaction is significant, but the latest data suggests the story is more complicated than a simple end to the AI rally.

The key question is whether investors are abandoning AI or simply becoming more selective about where the money goes. The difference could determine whether the next phase of the AI boom is driven by chips and data centers, software and inference, or a broader shift toward profitable AI applications.

Why did AI stocks suddenly fall?

The latest sell-off began after Anthropic CEO Dario Amodei urged AI companies to slow the pace at which they improve their models, citing risks including cyberattacks, misuse, and loss of control over advanced systems. OpenAI CEO Sam Altman agreed with Amodei, while xAI chief Elon Musk also supported the warning, creating a rare moment of agreement among major AI leaders.

Investors reacted quickly because the AI boom has been closely tied to expectations for continued spending on chips, servers, data centers and networking equipment. On September 14, the Philadelphia Semiconductor Index dropped about 5.2%, while Nvidia, AMD and Micron also suffered sharp declines as investors questioned whether AI spending could eventually slow.

The concern is not necessarily that companies will stop using AI, but that the pace of new investment could change. If companies decide they need fewer massive training clusters, suppliers that depend heavily on rapid infrastructure expansion could face slower growth expectations.

The timing also mattered because the market was already dealing with higher borrowing costs and rising Treasury yields. Reuters reported that the 10-year Treasury yield briefly moved above 5% during the same period, adding another source of pressure for technology companies that need large amounts of capital.

Altman also made an important distinction that received less attention than the headline about slowing development. He said that “pacing” does not mean stopping and indicated that AI progress would continue, suggesting that the debate is about managing the speed and risks of development rather than abandoning the technology.

That distinction matters because a slower pace of frontier-model development does not automatically mean a collapse in AI demand. The more important question is whether companies continue spending heavily on the infrastructure required to operate existing models and deploy new AI services.

Fun fact: The selloff rippled through the industry, where companies are increasingly relying on debt and circular financing to fund ambitious AI spending plans even as global borrowing costs, reflected in multi-year-high bond yields, continue to rise.

Could AI development slow without ending the spending boom?

Yes, AI development and AI infrastructure spending can move at different speeds because companies need computing power for more than training the newest models. Existing models still require chips, servers, networking equipment, and data-center capacity when millions of users and businesses access them.

Nvidia’s latest financial results provide a useful example of this continuing demand. The company reported $89 billion in quarterly data-center revenue for its fiscal second quarter, up 117% from the same quarter a year earlier, while total revenue reached $96.2 billion.

Nvidia also said its cloud-industry backlog had surpassed $2 trillion and estimated that capital spending by the five largest hyperscalers could approach $800 billion in 2026 and $1.3 trillion in 2027. Those figures are company estimates, not guarantees, but they show that major customers are still planning enormous infrastructure investments.

Morgan Stanley reached a similarly large estimate from a broader industry perspective. Its September 9 research said major cloud providers could spend more than $1.4 trillion on AI infrastructure in 2027, while also asking whether the enormous investment can generate sufficient returns.

This creates a major gap between the headline narrative and the underlying economics. A company could become more cautious about training increasingly powerful models while continuing to expand computing capacity for inference, enterprise tools, AI agents, and other commercial applications.

The distinction gives investors another metric to watch besides the pace of model development. If infrastructure orders remain strong while frontier development becomes more controlled, the AI market could be entering a different phase rather than simply moving toward an end.

Is the real AI problem the return on all that spending?

The bigger financial question may be whether AI companies can turn enormous capital spending into enough revenue and profit. Morgan Stanley’s research specifically highlights this issue, noting that the scale of AI infrastructure investment is forcing investors to consider whether the spending can produce attractive returns.

That matters because the AI boom requires much more than buying GPUs. Companies also need data centers, electricity, cooling systems, networking equipment, and specialized infrastructure, creating a massive upfront bill before many AI projects can generate meaningful returns.

The investment cycle also increasingly involves debt and other financing methods. Reuters reported that technology companies have been turning to debt markets to help fund AI and cloud expansion, making borrowing costs an increasingly important part of the economics behind the buildout.

Higher interest rates can change the calculation even when demand for AI remains strong. A project that looks attractive when financing is cheap can become harder to justify when the cost of borrowing rises, and investors demand faster or larger returns.

The question for investors therefore goes beyond whether AI is useful. They also need to determine whether the revenue generated by AI services can eventually justify the billions being spent to build the systems behind them.

That could become one of the most important tests for AI stocks over the next several quarters. Strong revenue growth from infrastructure companies can support the investment story, but sustained spending without clear evidence of profitable demand could make investors more cautious.

Fun fact: Shares of chipmakers Nvidia and Micron Technology, which both produce advanced semiconductors used in AI infrastructure, slid 3.4% and 5.3%, respectively, on the day.

Could power and data centers become the next AI bottleneck?

AI requires physical infrastructure, and electricity is becoming one of the most important constraints on how quickly that infrastructure can expand. Large data centers consume significant amounts of power, meaning AI companies increasingly depend on utilities, transmission networks, generation capacity, and construction projects.

The issue is already creating pressure outside the technology sector. Reuters reported that New York proposed requiring developers of new data centers to make community investments of at least $1 million per megawatt of utility demand, reflecting concerns about electricity use and the effect of data centers on local communities.

That means the future of AI spending may not depend entirely on whether companies want more GPUs. They may also need to determine whether they can secure enough electricity and physical capacity to operate those GPUs at the scale they have planned.

The infrastructure challenge can create opportunities for companies outside traditional technology stocks. Energy storage, power generation, data-center construction and other infrastructure businesses can all become part of the investment chain supporting AI.

It also introduces another possible limit to the AI buildout. Even if companies continue demanding more computing power, shortages in electricity, grid connections, cooling capacity or construction could slow how quickly new data centers become operational.

For investors, this makes the AI market broader than a group of chipmakers and software companies. The next phase of AI infrastructure could increasingly depend on the companies and industries responsible for supplying the physical resources that allow data centers to operate.

Could cheaper AI models change the AI stock story?

Cheaper AI models could change the economics of the industry by reducing the cost of delivering AI services. Reuters reported that competition from Chinese models such as Kimi, Qwen and DeepSeek is raising concerns that lower-cost alternatives could put pressure on the pricing of more expensive models.

Lower prices could have two very different effects on the market. They could increase adoption by making AI more affordable, but they could also reduce the amount companies can charge for individual AI services.

That creates a new question for investors: will AI companies make more money because usage expands, or will lower prices limit their margins? The answer could influence how much infrastructure companies need and how quickly they can recover their investment.

The shift from training to inference could also become important here. Training creates demand for enormous computing clusters, while inference creates ongoing demand as customers repeatedly use AI systems for searches, coding, business tasks, and automated workflows.

A cheaper model could therefore increase usage even if it generates less revenue per request. That could keep demand for computing infrastructure growing while changing which companies capture the most value from the AI ecosystem.

This is why a slowdown in frontier-model development should not automatically be interpreted as a collapse in AI. The industry could instead move toward cheaper models, higher usage, stronger competition, and a greater focus on making AI services commercially sustainable.

What should investors watch as the AI market enters its next phase?

The next phase of the AI market will likely be judged by several indicators rather than one dramatic headline. Investors will be watching capital spending plans, data-center demand, AI revenue growth, margins, financing costs, and evidence that businesses are actually getting value from AI.

The recent market reaction also showed that different parts of the technology sector can behave very differently. Semiconductor and infrastructure companies were hit particularly hard, while software and cybersecurity stocks helped major indexes recover from their worst levels during the September 14 session.

That difference could become more important if investors start looking beyond the amount of money being spent on AI. Companies that can demonstrate measurable revenue or productivity gains from AI may receive more attention than companies whose valuations depend primarily on expectations for future infrastructure demand.

Cybersecurity could also become a larger part of the story as AI systems become more capable. The same technology creating new productivity tools can create new security risks, potentially increasing demand for systems designed to monitor, protect and control AI-enabled applications.

There is also evidence that the physical AI buildout remains active despite the recent stock-market pressure. Reuters reported that companies such as CoreWeave and Super Micro raised expectations around AI infrastructure demand in August, showing that orders and capacity expansion had not suddenly disappeared.

The most important takeaway is that the current sell-off does not answer whether the AI rally is over. It shows that investors are beginning to examine the pace, cost, financing, physical limits, and potential returns of the AI buildout more closely, which could lead to a market that is more selective about where AI spending creates lasting value.

TL;DR

• AI stocks fell after Anthropic CEO Dario Amodei urged companies to slow the pace of AI development because of safety risks.
• The sell-off does not necessarily mean AI infrastructure spending is ending, as Nvidia and major cloud companies continue reporting strong demand.
• Investors are increasingly focused on whether enormous AI spending can produce enough revenue and profits to justify the investment.
• Rising borrowing costs, data-center power needs, cheaper AI models, and changing demand could reshape the AI market.
• The next phase of the AI rally may depend less on building bigger models and more on turning AI infrastructure into profitable products and services.

This article was made with AI assistance and human editing.

If you liked this, you might also like:

• Has Wall Street’s growing Bitcoin demand finally made BTC less dependent on China’s economy and liquidity cycle?
• Could China’s weakening credit impulse be the warning signal behind Bitcoin’s next major market setback?
• Why Bitcoin has remained surprisingly resilient as oil climbs above $90 and traditional markets come under pressure