New rules introduced by the US on Monday (16 November) have brought the term “remote computing” into the spotlight. What exactly is it? And how will the new rules affect the technology, artificial intelligence (AI) and semiconductor sectors in Singapore and the region?

The US Commerce Department announced new rules on Monday restricting exports of advanced AI chips made by US companies such as Nvidia to data centres in five economies: Singapore, Malaysia, Thailand, Indonesia and Brazil. The authorities said Chinese companies had been accessing computing power remotely through these data centres to train Chinese AI models.

The rules require a US government licence before advanced AI chips can be exported to these five economies, to prevent Chinese companies from using overseas data centres to circumvent US restrictions on chip exports to China.

Overseas data centres providing substantial computing capacity to companies headquartered in China must obtain a US export licence within 90 days to continue providing these services, or face further export restrictions. Licence applications will be reviewed on a case-by-case basis.

The announcement put many Singapore-listed technology and AI-related stocks under selling pressure, including chip-testing company AEM Holdings, semiconductor equipment parts supplier Frencken Group and semiconductor equipment manufacturer UMS Integration.

Lianhe Zaobao takes a closer look at remote computing.

1) What is remote computing?

Remote computing allows users to rent access to high-performance chips and computing resources at data centres in other economies, rather than spending heavily to buy or own AI chips or shipping them to the economy where they operate.

In Singapore, Singtel provides computing power through its RE:AI service. It buys graphics processing units (GPUs) and uses its data centres and related infrastructure to supply computing power to businesses and institutions. Singtel primarily serves customers in Singapore.

Singapore-based computing service providers that also serve overseas customers include Megaspeed, which has recently been embroiled in a controversy over an extramarital affair involving Jensen Huang and Huang Le, as well as HPC-AI Tech and Sustainable Metal Cloud.

There are currently no statistics showing which economies provide the most remote computing power. However, according to AI infrastructure analytics and market monitoring platform CUDO Observatory, the US currently leads by a wide margin, accounting for 75.3% of global operational computing capacity. China and Germany follow with 8.8% and 2.1% respectively.

2) Why has the US placed the five countries under export controls?

Although the US has banned exports of advanced AI chips to China, it had not previously prohibited Chinese companies from renting computing capacity from data centres in other economies. In other words, Chinese companies could still use the technology overseas to train AI models, even if the advanced chips never entered China.

3) How will the new rules affect Singapore’s computing services industry?

Phillip Nova’s Head, Market Insights, Priyanka Sachdeva, told Lianhe Zaobao that the new US rules could subject computing service providers to stricter compliance requirements and the risk of declining revenue.

“Higher compliance costs are almost inevitable. Companies may need to spend more on customer verification, documentation, legal reviews and monitoring how their technology is ultimately used.”

LIU Han-Wei, an associate professor at Singapore Management University’s Yong Pung How School of Law, also said computing service providers would have to conduct more checks on customers’ identities and end uses to comply with the new US rules. This could raise compliance costs and slow transactions.

In a joint statement on Tuesday (17 November), the Ministry of Digital Development and Information and the Ministry of Energy, Trade and Industry said the Singapore government was carefully studying the US Commerce Department’s new rules and engaging affected companies to understand the potential impact. The government was also in contact with the relevant US agencies to explore how to help Singapore companies secure a reliable supply of chips.

The two ministries also stressed that Singapore places great importance on its international reputation and will not tolerate any company using its association with Singapore to circumvent or violate other countries’ laws, including export controls, in cross-border business activities.

Responding to queries from Lianhe Zaobao on Tuesday, Singtel said it expected the new rules to have a negligible impact on the company. It said its GPU computing business was a standalone AI platform dedicated to serving Singapore customers, most of whom were from the public sector and research institutes. It does not engage in chip re-exports.

Ja Ian Chong, an associate professor of political science at the National University of Singapore, said in an interview that the US could introduce further rules after the Nvidia hearing concludes.

4) Will Singapore’s semiconductor industry be affected?

In recent years, Singapore has consolidated its position as a semiconductor hub through substantial investments by multinational companies, including Micron, UMC and Applied Materials.

Could increasingly tight US controls on trade in advanced chips also affect Singapore’s semiconductor sector, deterring foreign investment or prompting existing chip plants to relocate overseas?

Sachdeva said a single set of new rules would not be enough to undermine Singapore’s position as a semiconductor hub. However, the reputational impact should not be underestimated, especially as the US-China technology war intensifies.

She also did not expect chip plants to relocate for now, given the substantial capital investment, mature infrastructure and other requirements involved in chipmaking. But if similar controversies were to recur, multinational companies might delay investment or consider other economies instead.