Major AI players have unexpectedly signaled a slowdown in development, triggering a broad selloff across global AI and semiconductor stocks. However, some market observers believe this short-term sentiment shock is insufficient to reverse the underlying investment thesis for AI infrastructure.

On September 14, AI and semiconductor sectors broadly declined. In Asian markets, SK Hynix fell over 6% and Samsung Electronics dropped 4%. In US pre-market trading, memory chip stocks also weakened collectively, with Micron Technology down nearly 5%, Western Digital falling close to 4%, and SanDisk sliding over 5%.

The selloff follows a rare coordinated statement from AI industry leaders. Anthropic CEO Dario Amodei published a lengthy post on X last Saturday, urging AI companies to voluntarily slow the development of frontier models and announcing that Anthropic would unilaterally introduce independent third-party evaluation mechanisms.

OpenAI CEO Sam Altman subsequently expressed support, and xAI founder Elon Musk also posted that "Dario is right." The rare consensus among three AI leaders quickly dampened market sentiment, raising investor concerns about the AI capital expenditure cycle.

Yet some investors argue that "slowing development" does not necessarily mean a reversal of the AI capex cycle. Demand for computing power, data center construction, and supporting infrastructure such as electricity, storage, networking, and cooling remains intact. If model development pace moderates, it might give the industry more time to absorb previous investments and push forward AI application commercialization.

Where the Selloff Originated

In his post, Amodei warned that within the next 6 to 12 months, AI agents could gain the capability to "control the entire internet," with potential losses reaching hundreds of billions of dollars. Previously, a threat intelligence report released by Anthropic disclosed that its Claude model had been used for weapons development, cyberattacks, surveillance, and fraud activities.

These warnings quickly resonated across the industry. Altman stated that the risk of human extinction from AI is "unacceptable," while Musk publicly endorsed Amodei's views.

Once the news reached capital markets, the AI supply chain came under pressure first. Takayuki Miyajima, senior economist at Sony Financial Group, noted that weekend comments about slowing AI development would exert selling pressure on Japanese AI and semiconductor-related stocks, with Middle East geopolitical uncertainty further amplifying risk-off sentiment.

AI assets, already trading at elevated levels, are particularly sensitive to shifts in expectations. Charu Chanana, chief investment strategist at Saxo Markets Singapore, said valuations for AI and chip stocks are built on assumptions of robust demand and rapid technological progress, adding that "even the mere possibility of delays is enough to trigger profit-taking."

Why Slower R&D Does Not Equal Peak Compute Demand

However, another group of investors believes the market may be equating "slowing AI development" with "cutting AI infrastructure investment."

Billy Leung, investment strategist at Global X Management in Sydney, believes the three CEOs' support for a slower pace will not directly change investments in chips, electricity, or data center infrastructure. In fact, it may stretch out the entire development cycle.

"If commercialization and AI applications continue to grow while the pace of new capability development slows slightly, the industry could shift from 'spending on construction' to monetizing existing assets, which is essentially capitalization," Leung said.

Gary Tan, portfolio manager at Allspring Global Investments in Singapore, also believes the statements could create short-term pressure but are unlikely to derail the long-term AI investment theme. He pointed out that the AI industry remains in a relatively early stage, and given rapid technological evolution, other ecosystem participants may not be willing to slow down in tandem.

In other words, what the market truly needs to monitor is not whether AI research briefly decelerates, but whether AI capital expenditure experiences a substantial contraction as a result. If projects involving data centers, power, networking, and storage continue to move forward, the long-term logic for chip demand has not fundamentally changed.

The Real Test: Can Massive AI Investments Deliver Returns

Beyond the AI safety debate itself, the deeper concern in capital markets centers on whether the enormous AI infrastructure investments of recent years can ultimately translate into sufficient commercial returns.

Sebastien Mallet, portfolio manager at T. Rowe Price in London, said that the notion that AI "will change the world" does not guarantee that every related investment will generate attractive returns. As capital spending scales up, investors are shifting their focus from "how much computing power AI needs" to "how much money this computing power can ultimately make."

Meanwhile, strengthened safety regulation could also create new investment opportunities. Chanana believes areas such as cybersecurity and AI surveillance may see additional demand, while infrastructure companies in storage, networking, cooling, and power equipment could continue to benefit from projects already underway.

She noted that adding more safety safeguards will not make compute demand or AI applications disappear. Instead, it might make the industry's development path more prudent and sustainable.

Of course, dissenting voices exist. Investor Michael Burry, known for accurately predicting the 2008 US housing crisis, posted on X that the recent safety warnings in the AI industry may just be "hype and bluster," intended to mask real and uncontrolled growth deceleration.

For the AI industry, the core question has perhaps shifted from "whether AI will continue to develop" to whether massive AI capital spending can be sustained, and when compute investments will truly convert into profits. This will ultimately determine whether the current AI infrastructure rally is experiencing a sentiment-driven pullback or entering a phase of repricing valuations and business models.