Countries are being pressed to pick between two rival AI networks, and many are trying hard to avoid choosing just one. China’s World Artificial Intelligence Cooperation Organization (WAICO) was established in July 2026 with its headquarters in Shanghai and 29 countries on the founding agreement. The US-led Pax Silica initiative counts roughly two dozen partners and allies. The pressure peaked this month, when Reuters reported that the US was preparing to tell partners they must pick sides in the AI race with China.
U.S. News & World Report and several other outlets now rank the story among the biggest technology headlines of the week. The facts on the ground are messier than the headline suggests. Reporting describes overlapping initiatives, and at least one country, Kazakhstan, has joined both Pax Silica and WAICO. No outlet has shown that Washington has formally forced every partner to choose a single camp.
This analysis separates what is confirmed from what is still rumor, maps the countries in the middle, and looks at what the split means for chips, cloud contracts and model pricing. It builds on our earlier look at the AI infrastructure race, which argued that the hardware layer would decide who sets the terms. That argument now has a diplomatic layer on top of it.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Happened: Two Initiatives, Two Pitches
Two named international AI initiatives sit at the center of the reporting. WAICO is the China-backed intergovernmental organization. Pax Silica is the American answer, built around securing supply chains for semiconductors, computing capacity and other critical technologies. The two pitches differ in kind. Pax Silica sells security of supply to partners who already sit inside Western chip and compute chains. WAICO sells membership in a Chinese-hosted body that offers shared services and a seat at the table.
Reporting does not describe either as a treaty alliance in the Cold War sense. Pax Silica is described as an initiative supported by around two dozen partners and allies. WAICO is a formal organization with a headquarters city, a founding agreement and a July 2026 start date. That difference matters, because a looser initiative is easier to join alongside a rival one. Analysts at CSIS have framed the whole contest as one that should look beyond the US-China binary in the Global South, and the country-level evidence so far backs that view.
Pax Silica vs WAICO at a Glance
The table below sets out only what reporting has confirmed. Where details are missing, such as a full membership list for either side, the cell says so rather than guessing.
| Feature | WAICO | Pax Silica |
|---|---|---|
| Lead sponsor | China | United States |
| Type | Intergovernmental organization | Initiative with partners and allies |
| Start | July 2026 | Not confirmed in available reporting |
| Headquarters | Shanghai | None reported |
| Members | 29 countries signed the founding agreement | Roughly two dozen partners and allies |
| Core focus | AI cooperation and shared services | Semiconductor, compute and critical-tech supply chains |
| Reported dual member | Kazakhstan | Kazakhstan |
The takeaway from the table: the two bodies are not mirror images. One is built around supply security, the other around cooperation. Countries can reasonably find value in both, which is exactly what several of them are doing.
The “You Can’t Have It Both Ways” Warning
The sharpest line in the coverage comes from a Reuters-cited official, who said: “You can’t have it both ways.” It captures the American position in six words. A separate report attributed a similar warning to Washington, saying countries could not belong to both US- and China-led AI initiatives. That second report does not clearly establish who said it, so treat it as unconfirmed as either an official position or a verbatim quotation.
The gap between rhetoric and policy is the key detail. The available reporting says the State Department was preparing to issue such a warning. It does not say a final policy has been enacted, and no outlet has shown Washington formally requiring every partner to choose. Until a written policy appears, the “pick a side” language is best read as negotiating pressure aimed at partners who are hedging.
Who Is Caught in the Middle
The country list is where the “two blocs” framing starts to strain. Reporting names members on both sides, plus a handful that sit in both or split their spending. The table collects the country positions that published reporting confirms.
| Country | Reported position | Detail |
|---|---|---|
| Kazakhstan | Both | Joined Pax Silica and WAICO |
| Singapore | Pax Silica | Reported participant |
| Philippines | Pax Silica | Reported participant |
| Indonesia | WAICO | Reported founding member |
| Malaysia | WAICO | Reported founding member |
| Cambodia, Laos, Myanmar | WAICO | Reported founding members |
| Brazil | Split spending | About $444 million AI investment announced in August 2026, divided between Chinese and US technologies |
| Saudi Arabia | Mixed stack | Humain-m3 Arabic model reportedly built on a Chinese open-source system |
Kazakhstan: The Clearest Dual Member
Kazakhstan has joined both Pax Silica and WAICO, according to the reporting. It is the single best counterexample to the idea that a country must pick one network. Central Asian capitals have long balanced Moscow, Beijing and Washington, and The Diplomat asked in a September piece whether the region would be forced to pick a side. Kazakhstan’s answer, so far, is no.
Southeast Asia: Neutrality Under Strain
The split inside ASEAN is striking. Singapore and the Philippines participate in Pax Silica. Indonesia, Cambodia, Laos, Malaysia and Myanmar are reported as WAICO founding members. A regional bloc that prides itself on neutrality now has members on both sides of the line. Asia Times argued this month that the US-China AI race is turning ASEAN neutrality into a liability. If Washington does harden its “pick a side” stance, ASEAN will feel it first.
Brazil’s $444 Million Split Bet
Brazil offers the cleanest picture of hedging as national strategy. The country announced an AI investment of roughly $444 million in August 2026 and divided it between Chinese and US technologies. That is not an accident of procurement. It is a deliberate refusal to depend on a single supplier, and it arrives just as American officials are telling partners that dividing the bet may not be allowed.
Brazil also has the scale to make hedging work. A mid-sized economy can negotiate with both suppliers because each wants the contract. Smaller states have less leverage, and for them a formal American ultimatum would hit harder. The same logic applies to buyers across Latin America, Africa and the Gulf, where governments are weighing US chips against cheaper Chinese models.
Saudi Arabia’s Humain-m3 and the MiniMax Link
Saudi Arabia’s Humain launched Humain-m3, an Arabic-language AI model that is reportedly based on an open-source system from the Chinese startup MiniMax. This is how the blocs actually blur in practice. A Gulf state can still build a flagship national model on a Chinese open-source foundation. Rest of World described the pattern in a recent headline as the Global South merging US chips with Chinese open-source AI.
The reason is cost and access. Open-weight models from Chinese labs can be downloaded, fine-tuned and hosted on whatever hardware a buyer owns. A government that wants an Arabic or Swahili or Bahasa model does not need Beijing’s permission to use one. That is why we think open models are the leakiest part of any wall between the two camps. Our guide to the best open-source LLMs shows how many of the strongest options now come from outside the US.
Mazu and China’s Services Offer
China is pairing the WAICO charter with practical services. Mazu, China’s AI-powered meteorological early-warning system, is the leading example. China offered assistance to 30 countries to use it. Weather and disaster warning is a smart entry point. It is useful, low-controversy and hard for a recipient to refuse on political grounds, especially for coastal and storm-prone states.
The strategy contrasts with the American one. Pax Silica leans on supply-chain security, a promise that matters most to countries already close to the chip ecosystem. China’s pitch leads with tools that solve a visible local problem. For a government with a tight budget, a working forecast system can outweigh an abstract commitment about semiconductor supply.
What Public Opinion Says
Governments are not the only audience. Pew Research Center surveyed adults in 25 countries in 2025 on who they trust to regulate AI, and the numbers explain why so many capitals hedge. In Pew’s own words:
“A median of 53% of adults across 25 surveyed countries trusted the European Union to regulate artificial intelligence effectively, compared with 37% for the United States and 27% for China.”
Pew Research Center, How People Around the World View AI, 2025
Neither superpower wins a majority. The European Union leads both of them by a wide margin, which gives middle powers a political reason to look for a third path. Pew also found sharp country differences. On Indonesia, Pew wrote:
“In Indonesia, 64% of adults trusted China to regulate artificial intelligence effectively, making Indonesia one of the countries where trust in China was comparatively high.”
Pew Research Center, How People Around the World View AI, 2025
That fits with Indonesia’s reported role as a WAICO founding member. Public trust and diplomatic choice line up there. Pew also recorded strong home-country confidence elsewhere:
“In India, 89% of adults trusted their own country to regulate artificial intelligence effectively.”
Pew Research Center, How People Around the World View AI, 2025
The India figure points to a third instinct in the survey: national control over AI rules. Governments that feel that way tend to resist any external body, whether Chinese or American, that wants to write the standards for them.
Frontier Model Releases Raise the Stakes
The diplomatic fight is unfolding while the model race speeds up. Release trackers including Capital and Compute and NoCode MBA list a dense run of launches over the past six weeks. Anthropic shipped Claude Opus 5.5 on September 22 and Claude Sonnet 5.5 on September 28. OpenAI held its DevDay on September 29 and released GPT-6.1 Sol. Google made Gemini 4 Argon available in late September. Mistral opened a public preview of Mistral Large 4 on October 6. Moonshot AI’s Kimi K3.1 Pro appeared by October 23, and NVIDIA released Nemotron 3.5 models on October 15 and October 23.
| Model | Lab (home) | Release | Reported price per 1M tokens (in/out) |
|---|---|---|---|
| Claude Opus 5.5 | Anthropic (US) | Sept. 22, 2026 | Not listed in research |
| Claude Sonnet 5.5 | Anthropic (US) | Sept. 28, 2026 | $2 / $10 |
| GPT-6.1 Sol | OpenAI (US) | Sept. 29, 2026 | $2 / $10 |
| Gemini 4 Argon | Google (US) | Late Sept. 2026 | $2 / $10 |
| Mistral Large 4 (preview) | Mistral (France) | Oct. 6, 2026 | $0.68 / $2.09 |
| Kimi K3.1 Pro | Moonshot AI (China) | By Oct. 23, 2026 | Not listed in research |
| Nemotron 3.5 Bolt E4B | NVIDIA (US) | Oct. 23, 2026 | Not listed in research |
Our LLM leaderboard tracks how the top US models stack up against each other. On the Chinese side, our report on Kimi K3.1 Pro covers the open-weight angle that matters most to the countries in this story. Prices come from the trackers and should be read as reported figures, not official rate cards.
Market Impact: Chips, Cloud and Pricing
A hard “pick a side” rule would land first on the chip and cloud markets. Pax Silica is explicitly about semiconductors and compute capacity, so a country inside it is more likely to buy American accelerators and host American cloud regions. A country leaning toward WAICO is more likely to run open Chinese models on whatever hardware it can legally buy. Our coverage of NVIDIA B300 hardware spotted at China Telecom shows how contested that supply line already is.
Model pricing offers a second pressure point. The flagship US models cluster around the same $2 input and $10 output price per million tokens in the tracker data, while Mistral’s preview launched well below that and the open-weight Chinese options can run on a buyer’s own servers. A finance ministry weighing a national AI program sees that spread clearly. Buyers who pay for access to a US API are tied to US export and sanctions rules. Buyers who self-host open weights are not.
For US vendors, a pick-a-side rule is a double-edged sword. It could lock in the loyal partners, but it could also push hedging governments toward open Chinese models faster. For Chinese labs, WAICO gives a formal channel for distribution that their products did not have before July 2026.
Historical Context: Huawei, Export Controls and Non-Alignment
Washington has run this play before. In 2019 it added Huawei to its Entity List and lobbied allies to keep the company’s gear out of 5G networks. Allies split anyway. Some banned Huawei, some restricted it and some carried on. In October 2022 the US tightened controls on advanced chip exports to China, a step that pushed Chinese firms to build their own supply chain and pushed other countries to think about exposure.
The older precedent is the Non-Aligned Movement, founded in 1961, which gave states a way to refuse Cold War camps while still trading with both. Today’s hedgers reach for the same logic. The difference is that AI stacks run through chips, clouds, models and data rules, so staying neutral takes more engineering than it did in 1961. A country can stay out of an alliance. It cannot easily stay out of a cloud contract.
Foreign Policy’s October 7 report described the China-US AI race as playing out worldwide, while BCG has published an analysis of how the two powers are splitting the AI world. Caixin Global ran a commentary on September 1 calling the result an asymmetric fracture of global AI. The consistent theme across these outlets is that the fracture is real but uneven.
Three Paths: US, China and the European Alternative
The contest is not strictly two-sided. The Daily Star recently compared three regulatory paths, from the US, China and the EU. Mistral adds a European product angle. According to the trackers, Mistral reportedly planned to release open weights for Large 4 by the end of October 2026, which would strengthen Europe’s ability to offer an alternative to US- and China-dominated stacks. That plan is reported, not confirmed, so it is worth watching.
The Pew data from earlier supports the opening. With the EU trusted more than either superpower to regulate AI in the surveyed countries, a European third option has political room. It lacks the chip power of the US and the state-backed distribution of China. But if the open-weights release arrives, Europe becomes the easiest answer for a country that wants a model it can host without joining either camp. Our DeepSeek vs Claude price comparison shows how quickly cost gaps can sway buyers.
Five Predictions for the Next 12 Months
- Washington will soften the language, not the pressure. A formal rule banning dual membership would force a showdown with Kazakhstan and others. Expect carve-outs, side letters and sector-specific limits instead.
- More dual members will appear. With Kazakhstan already in both camps and Brazil splitting its spending, other mid-sized states will copy the approach before the pick-a-side policy hardens.
- Open-weight models will become the main battleground. The cheapest way for a government to avoid dependence is a downloadable model. Expect export-control debates to move from chips to weights.
- ASEAN will fracture further before it consolidates. With members on both sides, the bloc will struggle to hold a single line on AI standards.
- Europe’s open-weights move will decide the third path. If Mistral delivers Large 4 weights as planned, expect Gulf and Latin American buyers to cite it in negotiations with both superpowers.
These are our own projections, drawn from the pattern in the reporting. They are not quotes from officials or analysts.
What to Watch Next
The next signal comes from the top. The Star reported on October 23 that the US and China are racing to shape the AI agenda ahead of Xi-Trump talks in Shenzhen, citing sources. If AI rules end up on that table, the pick-a-side language could shift in either direction. A negotiated understanding would cool the pressure on hedgers. A failed round would raise it.
Three other markers are worth tracking: whether the State Department publishes the warning in writing, whether more countries sign WAICO beyond the 29 founders, and whether Pax Silica grows past its current two dozen. For deeper background on the models driving all of this, see our AI and machine learning coverage.
Frequently Asked Questions
What is WAICO?
The World Artificial Intelligence Cooperation Organization is a China-backed intergovernmental organization established in July 2026 and headquartered in Shanghai. A total of 29 countries signed its founding agreement.
What is Pax Silica?
Pax Silica is a US-led initiative focused on securing supply chains for semiconductors, computing capacity and other critical technologies. It is supported by about two dozen US partners and allies.
Can a country belong to both?
Yes, at least in practice today. Kazakhstan has reportedly joined both Pax Silica and WAICO. A Reuters-cited official said “You can’t have it both ways,” but the available reporting says the State Department was only preparing such a warning, not enforcing a final policy.
Which countries are in WAICO?
Reporting names Indonesia, Cambodia, Laos, Malaysia and Myanmar as founding members, along with Kazakhstan. The full list of 29 signatories is not in the available reporting.
Which countries are in Pax Silica?
Singapore, the Philippines and Kazakhstan are reported participants. The complete roster of about two dozen partners was not detailed in the sources we reviewed.
Is the world really split into two AI blocs?
Not fully. Reporting describes overlapping initiatives and countries active in more than one network, so the claim of two fully formed, mutually exclusive blocs is unconfirmed.
What is Mazu?
Mazu is China’s AI-powered meteorological early-warning system. China offered assistance to 30 countries to use it.
How is Brazil handling the choice?
Brazil announced an AI investment of about $444 million in August 2026 and divided it between Chinese and US technologies.
