Xi Jinping Just Walked Into Shanghai's World AI Conference and Called AI a "Public Good." Meanwhile, Trump Attacked China Over AI Access. Here's the Play.
For the first time in the conference's history, Chinese President Xi Jinping walked into the 2026 World AI Conference in Shanghai (July 17–20) and delivered the opening keynote.
His message was precise: "AI development should not be a solo performance by a single country, but a symphony of international cooperation."
On the same day, according to Reuters, Donald Trump delivered what multiple outlets described as an "unprecedented attack against Beijing" over US AI export restrictions during a prime-time national broadcast.
That's two superpowers laying out opposite visions for how AI gets built, distributed, and controlled. For indie builders and solo operators shipping products globally, that's not abstract foreign policy. That's a choice about which infrastructure, which vendors, and which geopolitical alignment your stack depends on.
What Xi said in Shanghai
Xi's keynote was not a casual conference speech. It was a structured argument about AI's future and China's role in it.
The framing: AI is a "public good," not a national weapon. Development should emphasize "open source, openness, collaboration and sharing." Countries should cooperate on "capacity building" to prevent "new historical injustices" in AI: specifically, wealthier nations controlling AI access and leaving developing countries behind.
The concrete move: China announced it would partner with international bodies and developing countries on AI cooperation, with emphasis on ensuring equitable access to AI capacity and preventing concentration of AI power in a single nation or bloc.
The implicit message: The US restricts AI access to China. China will be the advocate for global, open AI access. Choose sides.
The US response: restriction, not cooperation
Hours earlier, Trump delivered a broadcast address that attacked China's AI capabilities and the costs of US export restrictions. The reporting is thin on exact language, but the Reuters summary is direct: unprecedented, pointed, and clearly framed as a counterattack to China's open-AI positioning.
The US policy in 2026 is clear: restrict advanced AI models and chip access to China, maintain US technological dominance, and treat AI as a strategic asset.
China's policy announcement is equally clear: position itself as the pro-open-AI power and win influence with developing nations that are locked out of US AI access.
What this means for your infrastructure choices
If you're building a global product in 2026, you're making bets about which AI platforms are available in which regions, what data residency rules apply, and whether your chosen vendor faces geopolitical restrictions that might break your product.
Here's the landscape:
US-based vendors (OpenAI, Anthropic, Google, Microsoft):
- Excellent models and inference infrastructure
- Subject to US export controls: cannot legally serve customers in certain jurisdictions
- May face sanctions or restrictions if the geopolitical climate shifts
- If you build on them and you want to sell into China, India, or emerging markets, you hit a wall
China-based vendors (DeepSeek, Alibaba, Baidu):
- Open source and open access models coming
- Positioned as the "global public good" alternative
- Challenges: data privacy concerns for US/EU customers, different regulatory environment, potential US sanctions
EU vendors (Mistral):
- Independent from both blocs
- Smaller models but growing
- GDPR-compliant and transparent about data handling
The choice you're making when you pick a vendor is not just technical: it's geopolitical. And that choice limits where you can sell, whose data you can process, and what restrictions you might face two years from now when policy changes.
The honest counter-take: Xi's "public good" framing is a strategic move
Let me be direct: Xi's speech is a strategic positioning move to isolate the US on export restrictions and paint China as the pro-access, pro-developing-nations player. It's not charity. It's power.
If China's AI capacity gets less restricted by US policy, China wins. If the US gets isolated internationally as the "AI access denier," that's a win for China's diplomatic position. Framing open-source AI as a moral good is effective strategy because it IS a moral good by many measures, but that doesn't make it less of a strategic play.
Similarly, the US restriction policy is not a bug: it's strategy. Maintaining a technological gap gives the US an advantage. That's realpolitik.
For indie builders, the honest take is: both sides are playing to win. Neither is playing for your convenience. You need to choose which restrictions, which risks, and which opportunities you're comfortable living inside.
What I'd actually do if I'm shipping globally
Vendor-agnostic architecture. Abstract away the LLM layer. If you're currently hardcoded on OpenAI, you're betting the geopolitical climate won't shift in a way that cuts off access. I wouldn't make that bet. Build for swappability.
Audit your data residency. If you process user data from China, India, or EU, know the regulations. Know which vendors can legally operate in those regions. If you pick a US vendor and you want to sell in China, you're already out of the market. That's a revenue decision, not just a tech decision.
Plan for 18-month volatility. Geopolitical restrictions change faster than infrastructure migrations. If your product depends on a specific vendor getting access to a specific region, you're on borrowed time. Build redundancy into your stack: either multiple vendors or fallback inference strategies.
Track the conference cycle. Xi's Shanghai appearance signals China is serious about AI access politics. Watch the next major tech conferences (Davos, CES, China's annual AI conference) for shifts in positioning. When geopolitical strategy moves, vendor policy follows within 6-12 months.
If you're small, pick simplicity; if you're scaling, plan for switching. Solo operators and early startups can pick a vendor and move fast. By the time geopolitical restrictions bite, you've either won or pivoted. But if you're scaling past $1M ARR and you're dependent on a single vendor in a restricted region, you're making a strategic choice to exit that market or rebuild. Make it intentionally.
Author
Lukas
@lukcombinator