Playing both sides of the U.S.-China AI “Cold War”

retained text

Saudi Arabia celebrated the launch of what it calls the world’s most advanced Arabic-language AI model earlier this month. The technology behind it came from China even as the kingdom spends billions building its AI infrastructure with American companies.
The Arabic model, Humain-m3, is based on an open-source system from Chinese startup MiniMax. Humain, the state-backed company that commissioned the model, has struck major deals with Nvidia and other U.S. technology companies to build its AI computing capacity.
Saudi Arabia is not alone. From Latin America to Southeast and Central Asia, countries are buying American chips while adopting Chinese open-source models, splitting key AI projects between companies from two AI superpowers. The trend presents a growing complication for Washington’s competition with Beijing. Even countries investing heavily in American AI infrastructure are not necessarily choosing American technology for every layer of the AI stack.
“Many countries around the world, particularly in the Global South, want the best of both worlds,” Kyle Chan, a fellow at the John L. Thornton China Center at the Brookings Institution, told Rest of World. Many countries often do not attach the same security stigma to Chinese technology that Washington does, Chan said. “Ultimately, many countries around the world want to keep their options open and not be too dependent on any single source of technology.”
Staying in the middle
In August, Brazil announced an investment worth roughly $444 million into AI, which it effectively divided between Chinese and American technologies.
Around $255 million will go toward supercomputing infrastructure in Rio de Janeiro, developed in partnership with China’s HuaweiHuaweiHuawei is a Chinese technology company focused on mobile phones and telecommunications, and is seen as a poster child for China’s global tech ambitions.READ MORE and iFlytek. The rest, around $189 million, will go to a separate supercomputer that Brazil is currently tendering, but expecting to source from Nvidia.
“The strategy is not to depend on a single company, technology or country,” President Luiz Inácio Lula da Silva’s administration said in a statement, adding that the investments were intended to strengthen Brazil’s sovereignty over its data.
For most countries, technological sovereignty does not mean building an entire AI stack at home, Jesse Marks, a China-Middle East research scholar at the Australian National University, told Rest of World. Competing at the frontier has become prohibitively expensive, leaving governments to decide which technologies they can develop themselves and where they must rely on foreign suppliers.
For Saudi Arabia, it would be a “natural decision” to choose Chinese foundations for sovereign AI models if they perform better on relevant benchmarks, even if doing so carries political consequences with Washington, Marks said.
The U.S., however, has greater leverage in parts of the stack where its technology is harder to replace, such as advanced semiconductors, networking equipment, and sensitive cloud infrastructure. But Washington will struggle with countries that want AI but cannot afford American systems or software, Marks said.
“Outside of the network of American partners who are trusted or can afford it, it’s Chinese AI. It’s open models,” he said.
These technological choices are also becoming intertwined with countries’ broader diplomatic relationships. “AI is one of a number of portfolios that middle powers have to consider when negotiating with a partner like the United States,” Marks said.
Washington has accused Chinese labs of “free-riding” and training their models from frontier American AI models at an “industrial scale.” China, having rejected those accusations, has warned it would retaliate if Washington implements any curbs on Chinese AI companies.
China catching up
Chinese companies are ramping up efforts to compete with American companies on different layers of the AI stack.
In Egypt, Huawei has bid to build AI data centers for the government, according to a source-based report by Bloomberg, competing against a proposal involving Nvidia. A successful bid would give Chinese companies another foothold in a market where AI infrastructure has so far been dominated by American technology.
Southeast Asia is another region where China-U.S. competition over technological influence is increasingly apparent.
Washington launched the U.S.-ASEAN AI Spark initiative in July, offering financial and technical support to Southeast Asian governments interested in adopting American AI technologies for public-sector use.
Malaysia is considering going in another direction, evaluating Huawei’s Ascend 910C accelerators as the backbone of its sovereign AI initiative, Bloomberg reported.
The U.S. or China?
For now, the two AI superpowers appear to be taking different approaches to building influence abroad.
“China, like any leading power, wants the world to function in a manner that furthers its interests and replicates its own values,” Arindrajit Basu, a doctoral researcher in global governance of emerging technology at Leiden University, told Rest of World. “In the case of the internet and its allied technologies, this is a vision of governance providing a greater preference to the state in maintaining public order and security online. However, China is yet to engage in the naked coercion we have seen from the Trump administration that has reportedly clearly laid out an ultimatum.”
Washington is considering warning against countries trying to align with both U.S.- and China-led AI initiatives. That could put countries already engaging with both sides in a difficult position.
Kazakhstan, for example, has joined the U.S.-led Pax Silica initiative, which aims to build secure AI supply chains spanning semiconductors, compute, and other critical technologies. It has also joined China’s World Artificial Intelligence Cooperation Organization, or WAICO, which focuses on cooperation on AI development and governance. Uzbekistan signaled its intention to engage with both organizations last month.
Basu said the two initiatives currently have different aims.
Pax Silica is an invitation-only grouping focused on sharing critical resources among its members. WAICO, by contrast, is open to countries globally and presents itself as a governance body complementary to the United Nations. It is not currently tied directly to sharing computing resources or research.
“So, ideally China would want countries to choose, but I think it realizes that an ultimatum is unrealistic at this stage. Hence, we will see the more norms-oriented diplomacy that slowly but surely gets countries to use Chinese products and engage with Chinese capacity-building efforts,” Basu said.
Chan warned that Washington’s approach could instead encourage countries to preserve ties with both sides.
“Washington’s efforts to force other countries to pick a side will face growing pushback. Even close allies, such as Europe, are rethinking their dependency on American technology. Third countries will want to maintain flexibility and independence by working with both the U.S. and China,” he said.
Singapore could be an early test. A founding signatory of Pax Silica, the country has publicly acknowledged receiving an invitation to join the China-led AI organization and said it is considering the offer.
“We engage different countries through a range of multilateral and bilateral platforms,” Singapore’s Minister for Digital Development and Information, Josephine Teo, told Parliament last week. “The engagements arise at different times, serve different purposes, and evolve as our interests and circumstances change.”

document fetched 2026-09-17 02:49 · canonical https://restofworld.org/2026/ai-us-chips-chinese-open-source-models