The US just told 35 countries to pick a side in the AI cold war. Here's what's really going on.

Imagine receiving a letter from a powerful ally that says, more or less: choose us or choose them, because you cannot have it both ways. That is roughly what dozens of countries are about to get from the United States government, and the subject is artificial intelligence.

According to a draft State Department letter reported by Reuters, the US is preparing to tell friendly nations they must choose between the American and Chinese artificial intelligence ecosystems. The letter is addressed to the 35 countries that signed a June "AI Opportunity Statement." The language, apparently, is unusually direct for a diplomatic document. "To be part of everything is to be part of nothing," the draft reads.

To understand what is going on, it helps to know the name at the centre of this: Pax Silica.

What is Pax Silica?

Pax Silica is an initiative launched last year to lock down supply chains for AI models, semiconductors and critical minerals. The name is a play on "Pax Americana" and silicon, the stuff chips are made from. Washington's initiative has garnered support from around two dozen countries, including key allies like Japan, Australia, and South Korea.

Members of Pax Silica have access to shared investment opportunities in AI-related projects, while those who sign the AI Opportunity Statement have symbolically agreed on a "common purpose" and "shared vision" with the US. In plain terms: there is a carrot (investment access) and now, it seems, a stick (exclusion if you play both sides).

US officials drafted the letter to make clear that "you can't have it both ways," according to a US official who spoke to Reuters on condition of anonymity. The subtext, as one analysis put it, is that neutrality now reads, in Washington, as a form of defection.

What China has been building in parallel

The US letter does not exist in a vacuum. In July, Chinese President Xi Jinping launched the rival World Artificial Intelligence Cooperation Organization, which promotes China's open-weight AI models as an alternative to US systems from OpenAI and Anthropic. This organisation gives countries a way to access and build with Chinese AI, without going through American gatekeepers.

The pitch is partly about price. A fresh data audit of the US-China AI race finds American systems still ahead on benchmarks, while Chinese labs increasingly win the world on cost, with US capital and chips holding the frontier gap roughly steady. For a country with a tight government budget and no particular loyalty to Silicon Valley, that is a compelling offer. Hugging Face's chief executive put numbers on it this month, telling CNBC that China is winning the open-model race with 41% of downloads.

Kazakhstan's decision to join both Pax Silica and China's AI initiative has raised questions about the feasibility of maintaining relationships with both blocs. Kazakhstan is not an outlier. Many countries in Central Asia, Africa and Southeast Asia have deep economic ties with both Washington and Beijing, and they have no obvious reason to close off either relationship.

Why now?

The timing reflects a genuine shift in how policymakers in Washington view the AI competition. For a while, the assumption was that American companies were so far ahead that allies would gravitate towards them naturally. That assumption has worn thin.

Chinese open-weight models have closed much of the gap on the proprietary systems from OpenAI and Anthropic. China has also used its near-monopolies over critical minerals as a retaliatory weapon in a tariff war, prompting the US to ramp up efforts to source minerals domestically and from allies. Control over those minerals is, ultimately, control over who can build AI hardware at scale. The supply chain argument and the model argument are now fused into one.

There is also a domestic pressure angle. The enforcement machinery has been tightening in parallel, with the Commerce Department examining how China gets remote access to Nvidia chips it cannot buy, and the Treasury Secretary having threatened to sanction Chinese AI firms outright. Against that backdrop, asking allies to formally pick a side looks less like aggression and more like a logical next step to officials in Washington.

The case against the ultimatum

Worth pausing here, because the US government's framing deserves some scrutiny.

First, the document itself. Reuters could not establish when Washington will send the letter, or whether the text changes first. The draft is undated. The State Department said it would not comment on "purportedly leaked internal documents." So we are, at this stage, looking at a draft that may be revised or quietly shelved. It is worth treating the harder-edged claims with some caution until the letter actually lands.

Second, the framing of a binary choice ignores how most of the world actually operates. Countries in the Global South do not experience the AI race as a Washington-versus-Beijing drama. They experience it as a procurement decision: which tools are available, at what cost, and on what terms. Telling them they can only shop at one store is likely to produce resentment rather than loyalty, particularly when the American option often costs more.

Third, American industry is not united behind the closed approach either. Nvidia and several other large firms broke with OpenAI and Anthropic over open weights this year. So the US is asking other countries to commit to an ecosystem that its own tech sector is visibly divided over.

China's embassy in Washington pushed back on the US plans, saying such moves would slow global AI progress and serve no one's interests. On August 19, China called for respect for each country's "digital sovereignty," arguing that countries should be free to choose their technology partners according to their own needs. Beijing's messaging is, predictably, self-serving, but it lands well in capitals that bristle at being told what to do by either superpower.

Who actually gains and who carries the cost?

The clearest winners, if the initiative holds, are the large American AI companies. Locking in allies to the US ecosystem means more customers for OpenAI's APIs, more demand for Nvidia hardware, and more leverage over global AI standards. The geopolitical argument and the commercial interest point in the same direction, which is usually a sign that the commercial interest is at least partly driving the argument.

The countries being asked to choose carry most of the risk. A smaller nation that commits to Pax Silica and then finds Chinese models have become significantly better or cheaper has given up options it cannot easily recover. Conversely, any country that leans towards Beijing's orbit may find itself cut off from American cloud infrastructure, chip supply chains, and investment flows that it relies on. Neither option is comfortable.

Kazakhstan's dual membership illustrates the difficulty the US faces in persuading partner nations to limit economic ties with Beijing in line with US security priorities. Expect more countries to drag their feet, hedge quietly, or sign the statement while continuing to use Chinese models in practice. Pieces of paper rarely stop procurement decisions.

The bigger picture

The rivalry between the United States and China in artificial intelligence is taking on a new dimension. According to Reuters, Washington is considering asking dozens of countries to choose between the American and Chinese technology ecosystems. What this week's story signals is that AI has moved fully into the domain of foreign policy, trade agreements, and alliance management. It is no longer just a technology story.

That shift has consequences beyond diplomacy. When governments start treating AI models the way they treat weapons systems or telecommunications infrastructure, the incentives for the companies building those models change. Safety research, open publication, and international collaboration all become harder to justify when your government sees your work as a strategic asset that should not be shared.

Whether the letter lands softly or creates a genuine rupture with fence-sitting nations, the underlying dynamic it reflects is already real. The world's AI infrastructure is splitting along geopolitical lines, and the countries that did not build the technology are being asked to choose which version of it they trust. That is a question with no clean answer, and no shortage of people with an interest in how you answer it.

If this one was useful, there's plenty more on the site. Pieces on how AI works, plus coverage of AI news, the downsides included. All free to read, no account needed.

References

1. The US is about to make its allies pick a side in the AI cold war, The Next Web
2. The US is about to tell 35 countries to choose between it and China on AI, The Next Web
3. US Issues Ultimatum to Countries in AI Race: Choose Between Washington Coalition and China, ChinaTechNews
4. US to tell partners they must pick sides in AI race, ChinaTechNews
5. US wants to force partner countries to choose between Washington and Beijing in the AI race, The Decoder
6. US Tells 35 Allies To Choose Sides as Washington Threatens To Expel Anyone Joining China's Bloc, IBTimes UK
7. AI NEWS: Week of August 17, 23, 2026, Medium
8. US to tell partners they must pick sides in AI race with China, People News Today
9. Welcome to August 23, 2026, The Innermost Loop (Substack)