Is demand for AI technologies slowing down?

Shares in semiconductor companies posted strong gains over the past year, driven by investor bets on the pivotal role the sector plays in building the world's AI infrastructure.

However, a return of volatility to sector stocks has raised questions about whether it reflects broader concerns over slowing demand for AI technologies, according to CNBC.

Despite these concerns, a number of senior AI industry executives have ruled out any signs of a demand slowdown, stressing that companies have become more cautious about the costs of using these technologies — not about adopting them.

Pat Gelsinger, former chief executive of Intel and currently a general partner at Playground Global, said demand for AI is "almost unlimited."

He noted that energy availability represents "the only real constraint" on its continued growth, adding that the economic value AI can deliver as its capabilities expand is immense, and that its impact extends to virtually every sector.

A combination of factors has driven increased volatility in the shares of semiconductor companies and AI data centre firms in recent months, including Meta's announcement of its intention to lease out surplus AI computing capacity — a move that contributed to a sell-off in the sector.

Although Meta's share price rose following the announcement, the move raised questions about a potential surplus of AI computing capacity across the market. Elon Musk's xAI also leased out part of its surplus computing capacity during the current year.

In a related development, Samsung, one of the world's largest memory chip companies, forecast a major jump in profits, yet its share price declined — after having risen more than 360% over the past 12 months — as investors questioned whether the stock could sustain the same pace of gains.

Despite these developments, industry officials insist that demand for computing capacity and AI-supporting infrastructure remains strong. Mark Boroditsky, chief revenue officer at Nebius, said the company is seeing exceptional demand that exceeds its current capacity to fulfil.

He added that this situation has been ongoing for some time. Nebius operates data centres based on graphics processing units developed by Nvidia.

Andrew Feldman, chief executive of Cerebras Systems, argued that the decisions by Meta and xAI to lease out their surplus computing capacity represent an exceptional case and do not reflect broader market conditions.

Feldman said demand for computing capacity far exceeds available supply, adding that the sector is suffering from a shortage of data centres, as well as limited availability of many key components needed to expand computing infrastructure.

Cerebras Systems, which listed its shares on the stock exchange during the current year, is one of a group of semiconductor start-ups seeking to expand their presence in the data centre market and compete with Nvidia.

South Korean start-up Rebellions, backed by Samsung and SK Hynix, confirmed it is seeing strong demand for its AI solutions.

Chief executive Sungyeon Park said momentum in AI infrastructure investment remains substantial, expressing his belief that Meta and xAI leasing out their surplus computing capacity does not indicate that major technology companies are overspending on AI infrastructure.

Lumentum, a company specialising in photonics and optical communications solutions for data centres, confirmed that its production capacity is fully booked for the next 5 years, amid strong demand for its products used in building AI infrastructure.

Chief executive Michael Hurlston said Lumentum is expanding its production capacity as quickly as possible to meet anticipated demand over the next 5 years.

The company's shares have risen approximately 600% over the past 12 months, buoyed by investor appetite for companies providing solutions to key bottlenecks in the construction of AI data centres.

In contrast, another strand of the AI debate centres on the extent to which companies are willing to bear the costs of using these technologies.

The recent period saw what has become known as "tokenmaxxing," whereby some companies encouraged their employees to use AI tools as much as possible, regardless of outcomes or economic return, relying primarily on models developed by companies such as OpenAI and Anthropic.

However, companies are now more focused on achieving a clear return on their AI investments, particularly as the cost of advanced models continues to rise compared with the open-source models offered by companies such as DeepSeek and Alibaba.