Nvidia’s Booming A.I. Business Collides With U.S.-Chinese Tensions

In early August, Bhutan’s King Jigme Khesar Namgyel Wangchuck made a journey from his mountainous, landlocked nation to Nvidia’s headquarters in Silicon Valley, home to the cutting-edge AI chipmaker. The King spent two hours touring the facility, listening as Jay Puri, Nvidia’s head of global business, outlined how Bhutan could combine its abundant hydropower resources with data centers and Nvidia’s chips to create innovative AI systems.https://wordpress.com/

This meeting was just one of many Nvidia has held in recent years with world leaders—kings, presidents, and ministers. Many countries have since invested billions into developing supercomputers or generative AI, hoping to secure a competitive edge in what is expected to be the defining technology of the century.

However, officials in Washington are raising alarms that Nvidia’s global outreach could potentially strengthen adversaries. The Biden administration is now considering new regulations that would impose stricter controls on AI chip sales, using them as a diplomatic lever.

The proposed framework would allow U.S. allies to make purchases without restrictions, but adversaries would be banned, while countries’ quotas would depend on their alignment with U.S. strategic interests, according to four sources familiar with the plans who spoke on condition of anonymity.

These restrictions could disrupt Nvidia’s global expansion, which CEO Jensen Huang has dubbed “sovereign AI.” Huang has traveled extensively this fall, logging over 30,000 miles in just three months, and Nvidia expects more than $10 billion in sales from markets outside the U.S. this year.

Countries like Saudi Arabia, Malaysia, and Bhutan, which are keen on acquiring Nvidia’s technology, could soon find themselves caught in the rivalry between the U.S. and China over AI supremacy. To purchase Nvidia chips, these nations may need to cooperate with approved U.S. and European cloud service providers and ensure the technology doesn’t end up in China’s hands.

Huang has launched an intense lobbying campaign to influence the new regulations, but it may be too late for any changes.

“Every administration has made it clear that unless China changes its global stance, the U.S.-China tech competition will only intensify,” said Klon Kitchen, senior fellow at the American Enterprise Institute. “Companies like Nvidia should brace for more tightening.”

A Growing Wave of Interest

Back in 2019, Sasha Ostojic, a former Nvidia executive from what is now Serbia, learned that his home country was considering Chinese-made AI chips from Huawei to support its tech startups. He persuaded Serbian officials to visit Nvidia instead.

After seeing the potential of Nvidia’s offerings, Serbian leaders decided to invest in four supercomputers, worth about $500,000 each. By 2021, Serbia’s startup ecosystem had exploded, with the number of businesses growing from 200 to 800.

The release of OpenAI’s ChatGPT in late 2022 validated Serbia’s decision. The AI-powered chatbot became an overnight sensation, fueling a global surge in demand for Nvidia’s chips, which dominate 90% of the market. Many countries realized they needed to build their own AI capabilities, as services like ChatGPT were unavailable in some nations, including Saudi Arabia and Vietnam.

“The ripple effect was that countries saw AI as a critical technology they couldn’t rely on foreign powers for,” said Keith Strier, former Nvidia executive and now senior vice president at AMD, a rival chipmaker. “This sparked a massive wave of interest.”

Governments worldwide began to view AI as a vital utility, one that could not be ignored without risking economic and military disadvantage. Many also sought the security of keeping data within their borders to protect sensitive information.

In response, Nvidia set up a division to work closely with governments. Puri and Strier met with foreign leaders to understand their computing needs, while Huang inspired investments in data centers packed with Nvidia’s chips.

Governments, particularly smaller ones, had traditionally rented computing power from cloud giants like Microsoft and Google, avoiding the costs of software updates and semiconductor advancements. Now, many are taking the plunge into self-sufficiency.

Denmark is one such example. Although its leaders had initially shown no urgency, after using ChatGPT, they realized the power of AI and reached out to Nvidia. In partnership with the Novo Nordisk Foundation, Denmark invested $100 million in a supercomputer, unveiled in October with 1,528 AI chips weighing over 30 tons.

“We need innovation and a strong link between research and business,” said Peder Lundquist, CEO of EIFO, Denmark’s Export and Investment Fund. “To excel, supercomputing power is essential.”

AI Diplomacy in Action

As nations lined up to buy Nvidia chips, the Biden administration enacted rules to regulate the company’s global sales. Officials were concerned that countries with ties to China might use Nvidia technology to bolster China’s AI ambitions, potentially giving Chinese firms access to critical chips. As a result, some nations now require licenses for chip purchases.

The United Arab Emirates (UAE) was one of the first to request a license. In 2022, it allocated $100 billion for AI investments, aiming to build the region’s fastest supercomputer.

The U.S. saw this as an opportunity. For years, it had struggled to convince countries to stop using Huawei, citing cybersecurity risks and its ties to the Chinese government. Now, licensing requirements provided a way to offer Nvidia chips in exchange for stopping Huawei use.

In October, the UAE unveiled a supercomputer built with $30 million in Nvidia chips after agreeing to cut ties with Huawei’s technology.

Over the summer, U.S. officials began telling chip industry leaders that they were developing new AI policies. The goal was to simplify the licensing process while making it easier to build AI data centers abroad, while still maintaining the ability to track chips and prevent diversion to China.

A draft of the new rules is circulating among industry lobbyists. Under the proposal, U.S. and European cloud providers could receive faster approval for data center projects, while other countries would face more scrutiny and longer approval processes.

In response, Huang and Nvidia’s general counsel, Tim Teter, have raised concerns with the administration, arguing that the new rules could harm Nvidia. Other industry voices have expressed similar worries.

“We fully support national security and are happy to assist the government with any information it needs,” said Nvidia spokesman Ken Brown. “However, caps on mainstream data center products would be a significant shift that could hurt economic growth without reducing misuse risks.”

The rules are expected to be finalized soon, and Huang is lobbying to have them eased under a potential new administration.

As for the future, Gregory C. Allen of the Wadhwani AI Center at the Center for Strategic and International Studies notes that the U.S. is in an intense AI race with China. He believes the U.S. should ensure its companies remain on its side, fully aligned with American interests.

“We don’t want to hand our competitors Gatorade and Nike shoes in this race,” said Allen. “We need to keep them on our team, and only on our team.”