US to 35 Nations: Pick a Side in the Global AI Split

TL;DR

• Washington’s sending formal letters to around 35 countries demanding they pick either US-led or China-backed AI frameworks — not both.
• The Pax Silica initiative locks down supply chains for AI models, advanced semiconductors, and critical minerals in a winner-take-all alignment.
• Countries that hedge by participating in Beijing’s competing framework will be shut out of US chip exports, cloud infrastructure, and frontier model access.
• It’s the clearest push yet toward a bifurcated global AI ecosystem where geopolitical loyalty trumps technical merit.

Washington’s Ultimatum to 35 Countries

The US government is preparing to send formal letters to dozens of countries with a stark message: align with Washington’s AI governance and supply chain framework, or lose access if you also work with Beijing’s competing initiative. According to Reuters, the effort targets around 35 nations and is tied to the Pax Silica program, which aims to secure end-to-end control over AI models, advanced semiconductors, and the critical minerals that feed chip fabrication.

The letters represent one of the most explicit forced-choice moments in the escalating US-China technology rivalry. Countries that sign up for China’s framework will reportedly be excluded from the US-led coalition — meaning no access to cutting-edge GPUs, restricted cloud partnerships, and potential limits on deploying frontier AI models built on American chips.

“The U.S. is preparing to tell dozens of countries they must pick sides in the artificial intelligence race with China, warning they will be excluded from a U.S.-led AI coalition if they also sign up for Beijing’s competing framework,” according to the Reuters report. The policy push follows years of escalating export controls on advanced chips to China, but extends the battleground from hardware alone to governance, standards, and industrial strategy.

Why Pax Silica Rewrites the AI Playbook

This isn’t just about chips anymore. It’s about locking in a full-stack dependency — from the lithium mines in Chile to the data centers in Dublin to the model weights running inference in Mumbai. Pax Silica attempts to bundle hardware access, safety standards, and supply chain participation into a single geopolitical package, then force countries to pick one patron.

And it’s a bet that the US stack — Nvidia GPUs, hyperscaler clouds, and OpenAI-style frontier models — is valuable enough that most countries will choose Washington over Beijing when the ultimatum lands. But is it?

I’ve watched export controls tighten for three years, and this feels like the logical endpoint: a world where AI infrastructure is as politically charged as Cold War-era telecommunications. If you’re a government in Southeast Asia or Latin America, you’re now being told that working with Huawei on 5G or buying Chinese AI chips for university research labs could cost you access to the latest CUDA libraries, Azure AI APIs, or GPT-5 derivatives.

The stakes are highest for countries with deep economic ties to China but security dependencies on the US — think South Korea, which ships memory chips to Chinese fabs but hosts American troops. Or India, which wants to build its own AI industry but relies on US cloud giants and competes with China regionally. These aren’t hypothetical trade-offs. They’re procurement decisions that now double as foreign policy litmus tests.

Think of it like this: Pax Silica is the tech equivalent of telling a kid they can only attend one parent’s holiday dinner — Thanksgiving at Mom’s or Christmas at Dad’s, but if you show up to both, neither house will let you in next year. Except the kid is a sovereign nation, and the dinner is access to the computational backbone of the next decade.

Critics — and there will be many — argue that hard alignment demands risk fragmenting not just the internet but the AI research ecosystem itself. Multilateral AI safety cooperation becomes nearly impossible if Chinese and American researchers can’t share benchmarks, attend the same conferences, or collaborate on alignment research without triggering export violations. Smaller states get squeezed into a zero-sum game where flexibility dies and every partnership is scrutinized for geopolitical purity.

Some allies will quietly push back. European governments, for instance, have spent years trying to chart a third way on digital sovereignty — wary of both US surveillance capitalism and Chinese state control. Forcing them to pick a side undermines that project entirely. And countries in the Global South may resent being treated as pawns in a great-power rivalry, especially when China offers infrastructure financing and doesn’t lecture them about democracy.

How US-China Rivalry Reshapes AI Supply Chains

This move is part of a broader US effort to contain China’s rise in advanced computing and AI, building on years of export controls and alliances like the Quad and Chip 4. Over the past several years, Washington has imposed sweeping restrictions on advanced chips and lithography tools while pressuring allies like Japan, the Netherlands, and South Korea to align their export policies.

Pax Silica appears to extend this approach from hardware chokepoints to a comprehensive governance and industrial-policy framework. It’s an attempt to lock in a US-centric stack across semiconductors, AI models, and critical materials — essentially creating parallel supply chains that don’t intersect.

For AI companies and cloud providers, the implications are immediate. Where you can sell high-end GPUs, deploy data centers, and train frontier models will increasingly depend on which geopolitical bloc a country joins. Nvidia, AMD, and Intel face a world where export licenses aren’t just about chip specs but about whether a customer’s government signed the right diplomatic paperwork.

Chinese AI firms and chipmakers, meanwhile, are being forced to deepen domestic supply chains and court nonaligned countries as alternative partners. Huawei, Alibaba Cloud, and emerging players like DeepSeek will pitch themselves as the neutral option — no political strings, no forced alignment, just technology. Whether that pitch lands depends on how coercive Washington’s ultimatum feels and how competitive China’s homegrown AI stack actually is.

The semiconductor supply chain is already the most geographically fragmented industrial system on the planet — Taiwan makes the chips, Japan supplies the materials, the Netherlands builds the lithography machines, and the US designs the architecture. Pax Silica threatens to split that system in two, with each half trying to replicate the other’s capabilities while cutting off collaboration.

What Comes After the Letters Get Sent

Assuming the letters go out as planned, the first thing to watch is which countries sign immediately and which stall. Japan, Australia, and the UK will likely fall in line quickly — they’re already deep in US-led semiconductor alliances and have limited economic dependence on Chinese AI infrastructure. South Korea and Taiwan are more complicated; both are critical nodes in the chip supply chain but also export heavily to China.

The second thing to monitor is how China responds. Beijing could retaliate by restricting rare earth exports, tightening access to its domestic AI market, or accelerating its own exclusionary framework. If China makes its own pick-a-side demand, countries in Africa, Latin America, and Southeast Asia could find themselves trapped between two incompatible systems with no neutral ground left.

Third, watch the AI safety and governance conversations at the UN, OECD, and other multilateral forums. If the US and China can’t agree on basic interoperability standards, international AI governance becomes a joke. Every safety benchmark, every red-team protocol, every model card standard will fracture along geopolitical lines.

Finally, pay attention to the enterprise software layer. If countries align with the US framework, that likely means procurement rules favoring Microsoft, Google, Amazon, and OpenAI. If they go with China, it’s Alibaba, Tencent, and Baidu. SaaS vendors, cloud resellers, and AI startups will need to pick which market they’re building for — and accept that the other one might be permanently closed.

Source: Reuters