Xi pitches an 'AI for the Global South' line in Shanghai, with Phnom Penh and Bangkok on the dais

Chinese President Xi Jinping used a Shanghai forum on 17 July 2026 to call for deeper cooperation on artificial intelligence among developing countries, with the prime ministers of Thailand and Cambodia alongside him on the platform. The framing was deliberate. Beijing is offering a parallel track to the US-led conversation on AI safety and chip export controls, and it is doing so with two Southeast Asian governments whose compute and data infrastructure it has been quietly underwriting for years.

What Xi is selling is a softer version of the China model: shared standards, state-supported compute, and a development-friendly argument that the West is using "safety" and "security" as cover to lock poorer countries out of frontier capability. The pitch lands in a year when Washington has tightened controls on advanced semiconductors and the European Union has moved to require disclosure of training data. Beijing's answer is to convene its own constituency.

The Shanghai set-up

The forum sat inside the run-up to the Shanghai Cooperation Organisation summit, a setting Beijing has spent two decades converting into a venue for non-Western security and now technology diplomacy. Thai Prime Minister Anutin Charnvirakul and Cambodian Prime Minister Hun Manet both appeared, signalling that Phnom Penh and Bangkok are willing to be visibly aligned with Beijing's technology agenda in front of their own publics and their Western partners.

The substance matters more than the photo. According to Nikkei Asia's coverage, Xi told delegates that developing countries should not be left to import AI systems designed elsewhere, and should instead cooperate on standards, capacity building and shared infrastructure. The two Southeast Asian leaders echoed the line. Hun Manet has been particularly receptive; Cambodia is one of the most China-dependent economies in the region on infrastructure finance, telecoms and surveillance-equipment procurement.

Thailand is the more interesting case. Bangkok under Anutin has continued the long arc of Thai hedging, deepening trade with China while staying inside the US alliance network and hosting joint exercises. Endorsing Beijing's AI frame in Shanghai costs Thailand relatively little and buys it optionality in the inevitable negotiation over compute access, model weights and data localisation rules.

What Beijing is actually offering

"AI diplomacy" without delivery is rhetoric. The harder question is what compute, what data and what money Beijing is prepared to put on the table. Chinese state banks have financed telecoms backbone, undersea cables and 5G rollouts across the Mekong region; the next tranche will be data centres, large model training clusters, and the power grid to run them. Several Chinese cloud and chip firms have already established regional beachheads in Thailand and Malaysia.

The strategic logic is plain. If the West gates advanced GPUs and EDA tools, China's domestic accelerator stack becomes the default export offering to the developing world, with the Chinese government as standards-setter and financier. If model architectures diverge, Beijing can lock in interoperability now and define the regional norm. This is industrial policy with a diplomatic wrapper.

It also answers a domestic critique inside China. Chinese Premier Li Qiang has spent much of 2026 publicly pressing the country's big tech and AI sector to commercialise faster. An external constituency of buyers and standards partners is a useful way to validate that push, and to put pressure on Washington and Brussels by showing that the export market for Chinese AI capability is widening, not narrowing.

The counter-read from Washington and Brussels

Western capitals will read the Shanghai forum through a different lens. The concern in Washington is not the rhetoric but the plumbing: Chinese data-centre financing in places like Cambodia, where governance over data flows is loose, and where US intelligence agencies have already flagged Chinese-built telecoms infrastructure as a security risk. The Trump administration's revised chip-export regime, expanded in late 2025, treats the diffusion of Chinese AI capability as a national-security question; the Biden-era allies-coordination framework has been carried forward, with Japan and the Netherlands broadly aligned.

Brussels adds its own layer. The EU AI Act's general-purpose model obligations, phased in through 2026 and 2027, will force Chinese model providers serving European users into compliance or out of the market. Beijing's offer to the developing world can be read as the symmetric move: build a parallel market where EU-style transparency obligations do not apply, and where Chinese providers have a regulatory edge.

There is also a third reading, sceptical of both poles. Some Southeast Asian officials privately argue that the US-led chip regime has so far done more to entrench Chinese domestic capability than to slow it, by giving Beijing a clear industrial-policy target. If true, then the Shanghai pitch and Washington's controls are mutually reinforcing rather than opposing forces, and the contest will be settled by which side ships faster, cheaper compute.

Stakes for the region

For Thailand and Cambodia, the immediate gain is symbolic and political: a seat at a forum where the agenda is being set, rather than at the receiving end of rules drafted in Washington or Brussels. The medium-term gain is conditional on whether Beijing's AI offer comes with the same kind of strings attached to its Belt and Road infrastructure lending: concessional finance that becomes expensive once dependency sets in, and technology stacks that are hard to leave once integrated.

The risk is asymmetry. Cambodia has less leverage to renegotiate than Thailand, given the depth of its exposure across telecoms, ports and power. Thailand has more room to play both sides, but the room narrows if US export controls are tightened further and if Japanese and Korean equipment vendors are forced into compliance with stricter end-use checks.

For the wider developing world, the Shanghai forum matters less for what was signed and more for what it signals: that Beijing intends to position itself as the alternative pole for AI governance, not merely a participant in the Western-led one. Whether that pole has real compute behind it, or only slogans, is the question that the next twelve months will answer. Watch the data-centre announcements out of Bangkok, Phnom Penh and Kuala Lumpur; that is where this diplomacy becomes industry.

Desk note: Monexus is framing the Shanghai AI forum as an industrial-policy event with diplomatic packaging, not the other way around. The Western wire line emphasises security risk; the Chinese line emphasises development access. Both are partly true, and both are partly instrumental. We have weighted the developing-country perspective because that is the constituency the forum was built to address, and have flagged the counter-read from Washington and Brussels without endorsing it.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://t.me/nikkeiasia
• https://telegram.me/nikkeiasia