---

Discussion

Red Velhouse:

Let’s start with Washington’s actual position. Reuters reported in August that the State Department had prepared a draft message for 35 countries warning that participation in China’s competing AI framework could be incompatible with Pax Silica. The draft was not confirmed as sent. By October 25, the State Department publicly said countries could not be trusted Pax Silica partners while substantively advancing a competing vision of super intelligence. Omar, what boundary is Washington drawing?

Omar Seidren:

It is not a boundary around chatbots alone. Pax Silica covers critical minerals, energy, advanced manufacturing, semiconductors, data centers, and logistics. Washington is treating the AI stack as strategically connected: hardware, cloud capacity, models, data practices, and research relationships can all create dependence. The signal is increasingly explicit, but it is not yet a documented universal legal ban.

Ann Tofado:

That distinction is politically decisive. The draft described Pax Silica membership as a commitment, not a subscription. Washington wants aligned expectations, not merely attendance at meetings. But there is still no public exclusion rule, named country expelled, or finalized letter delivered to all 35 governments. This is pressure designed to shape choices before the policy machinery is fully visible.

Kate Burvish:

For governments and companies, ambiguity itself has a cost. If countries cannot tell whether dual participation could affect access to American chips, investment, financing, or projects, they may act cautiously without a formal prohibition. Companies may also hesitate to build around suppliers that could later face sanctions or export controls. Uncertainty changes procurement decisions early.

Red Velhouse:

The clearest test case is Kazakhstan. It joined the U.S.-backed Pax Silica initiative and then became a founding member of WAICO, the China-backed organization established by 29 countries. Why does Kazakhstan matter so much?

Ann Tofado:

Because it turns an abstract dilemma into a real overlap. Kazakhstan has critical-mineral importance, and the Associated Press reported a potential Nvidia-related investment deal worth up to 10 billion dollars for a proposed data-center valley. Yet Kazakhstan also signed WAICO’s founding agreement, and there is no public evidence it has been expelled from Pax Silica. Washington must decide whether dual membership is disloyalty or evidence that its coalition must accommodate independent states.

Kate Burvish:

For Kazakhstan, diversification is economically rational. It can seek high-end computing and investment from the United States while maintaining Chinese finance, infrastructure relationships, and diplomatic channels. Middle-income economies rarely want one supplier for every layer of a complex system. The benefit is bargaining leverage; the cost is incompatible rules, compliance burdens, and possible retaliation.

Omar Seidren:

Technically, choosing a side is fuzzier than the phrase suggests. A data center can use American accelerators, Chinese or American models, locally governed data, and software from several countries. The stack is modular, which makes separation difficult but creates hidden dependencies. Open-weight software does not erase dependence on electricity, networking, chips, cloud orchestration, or maintenance. There is no one-button AI alliance.

Red Velhouse:

Indonesia has joined WAICO and says it should be able to work with both American and Chinese systems. Singapore is considering WAICO participation while attracting American investment, including an Anthropic Asia office. Ann, are these countries rejecting bloc politics or negotiating harder within it?

Ann Tofado:

Mostly negotiating harder. Indonesia presents flexibility as sovereignty: it wants to select the best option rather than accept a geopolitical package. Singapore is balancing investment, security relationships, and its role as a technology hub. Neither is necessarily declaring neutrality. Both are saying national interests cannot be reduced to Washington’s preferred definition of trust.

Kate Burvish:

Their choices reflect practical development needs. Who will finance power generation, train workers, build connectivity, provide local-language applications, and absorb data-center risks? China often emphasizes affordability, deployment assistance, and financing. The American offer is stronger in frontier computing, private capital, and high-end infrastructure. For many countries, the question is capacity first and model rankings second.

Omar Seidren:

That is why the contest is bigger than model performance. Stanford’s 2026 AI Index recorded about 285.9 billion dollars in U.S. private AI investment in 2025, compared with 12.4 billion in China, although that comparison excludes much Chinese state-directed financing. Investment does not determine adoption by itself. Lower-cost, open-weight, easier-to-deploy systems can win users even when they trail on frontier benchmarks.

Ann Tofado:

Beijing is also building an institutional argument. WAICO is presented as an inclusive organization for global governance and cooperation. China can portray American restrictions as an effort to preserve a technological monopoly while offering countries a seat in a broader framework. WAICO’s effectiveness remains unproven, but its diplomatic purpose is clear: make Chinese participation look like sovereignty and American restrictions look like exclusion.

Kate Burvish:

China’s influence also comes through deployment. AidData research cited by the Associated Press found Chinese official grants and loans supporting AI-related projects in at least 77 low- and middle-income countries. That does not mean China supplied a frontier model or complete stack to each country. But financing networks, data centers, public-service systems, and digital infrastructure creates relationships that persist after the original demonstration is forgotten.

Omar Seidren:

The less glamorous components often decide what survives: reliable power, cooling, networking, software maintenance, data governance, and skilled operators. AI systems are industrial projects with very large electricity bills. Whoever builds the physical and human infrastructure may shape the market longer than whoever gives the most impressive model demo.

Red Velhouse:

Saudi Arabia shows another kind of overlap. Its Humain platform reportedly uses an open model from a Chinese AI company while the kingdom also has an Nvidia computing agreement. Is that what Washington wants to prohibit?

Ann Tofado:

We cannot assume that. Using a Chinese model and buying American hardware is not equivalent, on the available evidence, to joining a rival intergovernmental initiative. The uncertainty is what Washington will count as substantive participation: formal membership, financing, model deployment, research ties, or some combination. If every commercial interaction becomes disqualifying, the policy is extraordinarily broad. If institutional commitments are the focus, it is narrower and easier to manage.

Kate Burvish:

Broad rules could raise costs for American firms as well as foreign governments. If customers must reject cheaper models or alternative suppliers to remain trusted partners, they may delay projects or build less capacity. Dual sourcing creates interoperability, cybersecurity, data-governance, and compliance costs, but forced exclusivity eliminates competition and bargaining power. That cost eventually reaches public budgets or consumers.

Omar Seidren:

Restrictions intended to create a clean boundary may encourage countries to build more modular systems so they can switch components. That makes the ecosystem harder to control. Modularity is not frictionless, though: a model trained under one data regime, running on another provider’s hardware, with a third country’s security requirements, can produce an interoperability swamp. The system works—until the rules change.

Red Velhouse:

Could incompatible export controls and data rules eventually make hybrid systems too expensive or insecure?

Kate Burvish:

Certainly. Diversification may require separate compliance teams, duplicated infrastructure, different standards, and safeguards against data crossing jurisdictions. But full alignment also has an opportunity cost: rejecting affordable suppliers or financing can slow development. The right choice depends on which layer is being protected and how much duplication a country can absorb.

Omar Seidren:

The main chokepoints may not be models. Advanced chips, cloud capacity, energy, critical minerals, networking equipment, and skilled labor are harder to replace quickly. Software can sometimes be swapped; an entire power system or semiconductor supply chain cannot. That is why Pax Silica’s broader scope matters: it is organizing dependencies before they become irreversible.

Ann Tofado:

But neither side controls the full stack. The United States has advantages in frontier companies, capital, advanced computing, and alliances. China has domestic scale, state-backed finance, open-weight systems, and infrastructure relationships. Their ecosystems remain interdependent in manufacturing, talent, hardware, and markets. Political rhetoric points toward separation while the technology resists a clean split.

Red Velhouse:

So is Washington creating a genuine security boundary, or turning commercial advantage into exclusivity?

Ann Tofado:

Both interpretations have evidence. Washington reasonably sees advanced AI as connected to military capability, economic power, and strategic dependence. But trusted-partner language also converts a security concern into a coalition test. Until the United States publishes membership rules or enforcement mechanisms, this is escalating political pressure—not a universal legal prohibition.

Kate Burvish:

For middle powers, the likely response is selective alignment: American high-end computing, Chinese infrastructure finance, local data controls, and multiple model providers. That can preserve choice if countries have enough technical capacity to manage the interfaces. Otherwise, modular sovereignty becomes a costly procurement problem.

Omar Seidren:

The next proof point is deployment. If WAICO produces funded projects, shared standards, or reliable access to useful systems, it becomes more than diplomatic architecture. If Pax Silica delivers resilient supply chains and investment, it becomes more than a trust label. The future will be decided by who can make the infrastructure work on Tuesday morning.

Red Velhouse:

The unresolved issue is whether Pax Silica’s trust standard becomes a binding exclusivity regime, a flexible case-by-case screen, or mainly a warning to countries testing both ecosystems. Watch for delivery of the reported draft message, formal Pax Silica membership rules, WAICO’s first funded projects, and sanctions or export-control cases involving dual participation. Kazakhstan, Indonesia, Singapore, and Gulf technology deals will show how much room middle powers still have to hedge. Sources and references for this discussion are available with the episode at Factolio.com.

---

Sources and References

These sources supported the factual material used in this discussion. Factolio’s panel discussion is AI-generated from researched evidence and is written in original language.

1. Reuters, republished by Investing.com — Exclusive-US to tell partners they must pick sides in AI race with China (NEWS)
2. Associated Press — Why some countries struggle to pick sides in the US-China AI race (NEWS)
3. U.S. Department of State — Pax Silica (PRIMARY)
4. State Council Information Office of the People’s Republic of China / Xinhua — 29 countries sign agreement on establishing World AI Cooperation Organization (PRIMARY)
5. Indonesia Coordinating Ministry for Economic Affairs — Indonesia Resmi jadi Founding Member WAICO (PRIMARY)
6. Brookings Institution — Middle power AI agency: Preserving choice between the United States and China (ANALYSIS)
7. Stanford Institute for Human-Centered Artificial Intelligence — Economy | The 2026 AI Index Report (DATA)
8. AidData, William & Mary — China Research by AidData (DATA)
9. Associated Press — US Treasury proposes a rule to crack down on high-tech investments in China, other adversaries (NEWS)
10. Stimson Center — America’s Middle Power AI Export Challenge (ANALYSIS)
11. Associated Press — Global AI safety concerns loom for meeting between Trump and Xi (NEWS)