Kazakhstan signed two agreements that, in theory, shouldn't be able to coexist. It joined Pax Silica, the technology alliance Washington built to secure the supply of semiconductors, critical minerals, and artificial intelligence infrastructure. And it also joined the World Artificial Intelligence Cooperation Organization, the bloc Xi Jinping launched in July as a direct response. Two memberships. One country. Zero apparent problem from Astana. The problem showed up in Washington.

Pax Silica is a technology cooperation agreement between the United States and its allies aimed at controlling AI supply chains, because whoever dominates critical minerals and semiconductors dominates who gets to train frontier models. That's the operative definition, no frills. Which is why Kazakhstan's dual membership isn't an administrative footnote. It represents the first visible crack in an attempt to build two mutually exclusive technological ecosystems in a world that ran on shared supply chains.

The U.S. diplomatic response, according to reports, involves notifying the thirty-five signatory countries of the "AI Opportunity Statement" that they can't play at both tables. Pick a side or lose out on the benefits of Pax Silica. The phrase circulating in diplomatic circles, don't play at two tables, has the dry tone of a corporate ultimatum dressed up as foreign policy.

Why does the United States care so much whether a country like Kazakhstan holds dual membership? The answer has to do with minerals, not ideology. Kazakhstan produces uranium, chromium, and other minerals critical for semiconductors and batteries. If that country keeps simultaneous access to both blocs, the logic of exclusivity underpinning Pax Silica collapses, because the alliance's purpose isn't just technical cooperation but also the management of scarcity.

Washington and Beijing converge on state control of frontier AI even as they compete on everything else. That convergence produces a paradox. Two powers that present themselves as opposite poles end up restricting access and forcing binary loyalties onto third countries that would rather not choose. Kazakhstan isn't acting rebelliously. It's doing what any rational state would do when faced with two offers that don't naturally exclude each other: take both.

The pattern isn't new. Something similar happened with Ceuta in 2021, when Morocco used migration control as diplomatic leverage against Spain. The difference here is scale and terrain. Instead of physical borders, we're talking about borders of access to compute, AI models, and minerals. A bigger player pressures a smaller one using a resource the smaller one controls, and the smaller one discovers it has more room to maneuver than the bigger player expected.

The Greeks invented soft power two thousand years before anyone coined the term, something I explore in Stones Don't Lie. They didn't conquer territories. They conquered minds by offering a model of civilization so attractive that local elites adopted it voluntarily. Pax Silica is trying to pull off the same move with infrastructure instead of gymnasiums and theaters. China offers an equally seductive version of the same package, and countries like Kazakhstan, Indonesia, or Nigeria have zero ideological incentive to choose between Sam Altman and his Chinese counterpart if both are offering investment, compute, and trained models.

What does the United States actually gain if it manages to impose exclusivity? Restricting the spread of open Chinese models that compete on cost and accessibility against the closed products of OpenAI and Anthropic. The openness of Chinese models — think DeepSeek and its successors — represents a structural threat distinct from traditional military competition. It's not about who has the most powerful model but about who provides cheaper access to emerging economies that can't afford high-value American licenses. I already explored this tension in the context of the conflict between the Pentagon and Anthropic, where state pressure on private AI actors reveals how much control Washington exerts even over the companies it claims to be protecting from the Chinese regime.

The one gaining power from this ultimatum isn't just the United States as a state but a specific handful of companies that need the global AI market to stay fragmented into two closed blocs. If open Chinese models circulate freely through emerging economies, OpenAI's and Anthropic's competitive edge erodes fast, because their business model depends on artificial scarcity rather than sustainable technical superiority. Cui bono. The eternal question. And the answer points to whoever already holds the power they want to keep.

When a structure forces a choice between two mutually exclusive alternatives, people start looking for third options the original design didn't account for. It happened during the Cold War with the Non-Aligned Movement. It's happening now much faster, because the costs of switching technological blocs are lower than the costs of switching military blocs. Kazakhstan doesn't need to build a parallel army to maintain both memberships — it just needs to keep signing papers and hoping neither giant actually follows through on its threats.

What does this mean for mid-sized countries that have neither critical minerals nor their own compute infrastructure? They get stuck in the worst possible position, without Kazakhstan's bargaining power but under the same pressure to pick a side. That's the silent cost of this hegemonic competition — not the fragmentation of the powers, but the shrinking of options for everyone else.

I have no certainty about how this ends. It could be that the United States manages to impose exclusivity and fragment the world into two closed technological ecosystems, replicating the bloc pattern we already saw in the Cold War but with silicon instead of ideology. Or it could be that the pressure on countries like Kazakhstan simply accelerates a third way — coalitions of countries that refuse to choose and end up negotiating their own space, the way India or Indonesia did during decades of non-alignment. What does seem clear is that the rhetoric of not playing at two tables reveals more anxiety than strength. If Pax Silica were as superior as Washington suggests, no one would need to threaten anyone to make them stay.

What alternatives will countries build if they'd rather not align fully with either bloc?

Sources:

1. Diplomatic reports on U.S. notifications to signatories of the "AI Opportunity Statement"

2. Announcement of the World Artificial Intelligence Cooperation Organization, July (Xi Jinping)

3. Records of Kazakhstan's accession to Pax Silica and WAICO

4. Author's previous analysis: "State AI: Washington and Beijing Converge"

5. Author's previous analysis: "Pentagon vs. Anthropic: Military AI Without Ethical Limits"