The AI arms race between the US and China is heating up

Tech billionaires aren’t the only ones fighting for control of AI. The US and China are too.

The two world superpowers are competing with each other for everything from the intellectual know-how to design AI hardware and software to the raw materials that power artificial intelligence systems. Both are also using government subsidies to spur new advancements.

Where the US holds a commanding lead at the moment is the development of generative AI systems such as large language models (LLMs), according to Frank Long with Goldman Sachs' (GS) office of applied innovation. These models hoover up existing data and use it as the basis for chatbots like Open AI's ChatGPT.

Another advantage for the US is that it can place export restrictions on high-performance semiconductors designed by companies like Nvidia (NVDA) that are in intense demand across the AI world. That, for now, is keeping development of the most sophisticated LLMs out of China’s reach, Long said.

China, however, is countering with its own maneuvers. It is restricting the export of chipmaking metals gallium and germanium to the US while also reportedly amassing a new $27 billion chip fund to back its own major projects.

The fierce competition for AI talent between the two countries turned up a notch this past week when the US Justice Department unsealed an indictment charging a Chinese national and former Google (GOOG, GOOGL) AI software developer with stealing 500 files of confidential code that the tech giant uses for its supercomputing data centers to train LLMs.

The government alleged that after stealing the intellectual property the defendant began simultaneously working for rival companies in China.

The charges “are the latest illustration of the lengths affiliates of companies based in the People’s Republic of China are willing to go to steal American innovation,” FBI Director Christopher Wray said in a statement.

'Horse race'

The fight for global supremacy of AI was a hot topic of discussion earlier this month at the Web Summit’s annual technology conference in Doha, Qatar, an event that attracted investors and tech executives from around the world.

AI leaders from the public and private sectors acknowledged that the US and China currently have the edge. The countries currently rank No. 1 and No. 2 in Tortoise Media's Global AI Index, which measures nations based on AI investment, innovation, and implementation.

But the leaders also said at the summit that it's too soon to know which countries over time will leverage the technology to the greatest economic and societal advantage. Singapore, for example, is now No. 3 in Tortoise's index after moving up rapidly in recent years.

"I think it's not going to be as straightforward of a horse race — this person or that person, this country or that country," Long said. "It's going to be a full stack with participants in competition" for the energy, computing power, data, and models needed for AI systems.

Long and his team at Goldman said in a recent white paper they suspect certain geopolitical “swing states” — like the United Kingdom, the United Arab Emirates (UAE), Israel, Japan, the Netherlands, and South Korea — may be best positioned to tap the technology and form AI alliances.

Other Asian countries beyond China already have several advantages. Taiwan is home to a cutting-edge semiconductor manufacturer, Taiwan Semiconductor Manufacturing Company (TSM), which produces 90% of the world’s most advanced semiconductors and 68% of semiconductors worldwide.

Japan and South Korea are also home to leading semiconductor manufacturing and design firms, and they are setting aside more government funds to drive AI advancement.

Japan allocated $13 billion to the technology in its 2023 budget, up from $8.6 billion in 2022, while South Korea committed $470 billion over the next 23 years to create the world’s largest semiconductor manufacturing hub.

Other potential AI hubs could emerge in Europe and the Middle East. The Netherlands — home to ASML (ASML) — is already the world’s sole manufacturer of ultraviolet lithography machines, required to fabricate leading-edge semiconductors.

The UAE has a $10 billion fund to invest in late-state technologies, while Israel has also attracted billions in private AI investments.

Global power will shift towards nations that produce versus solely consume AI technology, according to Alaa Abdulaal, head of digital foresight for the Riyadh-based multilateral foundation Digital Cooperation Organization (DCO). The group's mission is to recommend government policies that promote access to technology.

"It cannot be only done by government itself," Abdulaal said. "It needs to take a cooperative approach, where we have at the same table the private sector, public sector, civil society — all of them sitting together to come up with the right set of frameworks for AI."

'Getting much closer'

Today, most countries have adopted, at minimum, a national strategy to develop and protect against the potential risks of AI. And some have cleared the way for regulations, though none have been tested against market forces.

In March, the EU is expected to adopt new legal restraints around AI. Earlier this year the bloc's member states signaled their agreement with the AI Act, the world’s first comprehensive legislation to regulate the technology.

Federal laws specific to AI don’t exist yet in the US or UK, and it’s unknown whether that will happen.

In October, President Biden issued an executive order to encourage safe AI development, including privacy protections. More than a dozen US states have adopted multiple AI-related laws.

The UK, for its part, rolled out a “National AI Strategy” in 2022 and committed roughly $4 billion in chip development subsidies. The nation has also attracted the third-largest level of private investment in AI, behind the US and China.

State subsidies and export bans may not hold sway forever as nations compete, said Jihad Tayara, CEO of Evoteq, a Dubai-based company that facilitates public-private collaborations to build AI into public infrastructure.

"It's getting much closer," Tayara said of the digital divide, noting that most nations have access to high-speed 5G data capabilities.

And human capital, along with access to open-source models, will matter too.

"Today we're seeing open-source models emerging that are in many use cases equivalent in capability to the absolute most frontier models," Long said.

Nations that prepare their workforces to develop and implement AI models will gain a considerable edge, Abdulaal said.

"This opportunity cannot be seen unless we have the right people in place."