Symphony, not a solo

The EU can cooperate with China on open-source solutions, and treat AI as a means to achieve sustainable solutions rather than an end in itself

At the opening ceremony of the 2026 World Artificial Intelligence Conference and High-Level Meeting on Global AI Governance in Shanghai, President Xi Jinping said that "AI development should not be a solo performance by a single country, but a symphony of international cooperation". And the agreement on the establishment of the World AI Cooperation Organization was signed during the meeting.

This will address a real challenge. The race for tech sovereignty in the AI sector is fragmenting the global AI landscape to unprecedented levels, to the detriment of innovation and access, especially for developing countries.

AI keeps moving at breathtaking speed, with ever more powerful models topping the charts every week. US labs still lead in multiple aspects, but Chinese open-weight models — DeepSeek, Qwen, GLM and Kimi — have their own edges. In the United States, top labs announce mind-blowing compute investments while companies such as OpenAI and Anthropic chase billionaire initial public offerings to monetize their lead. Four companies have now crossed a $4-trillion valuation, a figure without precedent. The US is rolling back regulation, blocking states from legislating and fast-tracking data centers that increasingly need their own power based on the conviction that it is engaged in an existential race for AI leadership with China.

Michael Kratsios, the US administration's top AI official, has spoken openly of "promoting the export of full-stack US AI technology packages to allies and partners worldwide". At the same time, it is likely that Washington tends to weaponize other countries' AI dependence — on models, the cloud and the advanced GPUs that mostly come from Nvidia. The US' measures on non-US users access to Anthropic's Mythos and Fable 5 have only pushed other countries to double down on their tech-sovereignty plans meant to reduce reliance on single foreign suppliers.

The European Union feels this acutely. Despite years of regulation to force competition and interoperability, dependence on big tech has stayed almost the same. Amazon, Microsoft and Google own almost the entire cloud market; Alibaba gets the crumbs, and European operators barely register.

The European Commission's first answer was to build its own compute — aware that at least 70 percent of global data-center capacity sits in the US and China. In recent years, the EU has announced 19 AI factories, with at least seven "gigafactories" of at least 100,000 advanced GPUs each on the way. But this simply swaps dependence on hyperscalers for dependence on Nvidia: Almost 90 percent of these factories have chosen Nvidia GPUs. This is hardly a win for the EU's tech sovereignty.

The deeper problem is that Europe is chasing the US model — data-hungry, energy-hungry, water-thirsty infrastructure built in the obsessive pursuit of artificial general intelligence, however vaguely defined. Is that what Europe needs? Probably not. European countries don't need the most power-consuming AI; they need stacks fit for real use cases in transport, health, energy, manufacturing, government and finance. For each, the optimal solution rarely looks like a one-size-fits-all gigafactory. Emulating the US model with less money is like "chasing Icarus with smaller wings".

The EU should work together with China to steer AI toward the public interest, the global good and the benefit of humankind.

China offers Europe a more attractive template. Whatever US commentators claim, Beijing is not primarily racing toward AGI. It is leaning on cheaper, more sustainable compute — less advanced GPUs, ordinary CPUs — and deepening its use of open standards to develop solutions that are fit for purpose. Often that means smaller models on smaller data, running centrally or at the edge, sparing needless costs in energy, training, inference and fine-tuning.

This is what Europe, a global industrial powerhouse, should do. The European Commission's Apply AI Strategy points toward this goal — but it comes with almost no dedicated funding. The EU's Tech Sovereignty Package, published on June 3 — comprising two legislative proposals and two strategic frameworks — promises to boost European data capacity at scale in the form of an open-source stack offering secure, high-performing solutions across sectors. If it works, Europe could emerge as a more sovereign AI power, wrapped in a regulatory framework built for fundamental rights, competition, interoperability and sustainability — now being overhauled in the name of genuine simplification.

Will the plan work? It may depend on how the EU adapts its China policy in the face of reality: Europe and China are deeply complementary. Europe depends on China for rare earths and much else; China needs the European market and technologies such as ASML's lithography for advanced chips. Duplicating such unique expertise would waste enormous resources.

No single country can set AI rules for the whole world. It is significant to promote EU-China collaboration that might be exactly the step toward a fairer, more sustainable and more sovereign AI future. The two sides are hardly "like-minded", yet they share strikingly similar perspectives: open-source solutions and the vision of AI as a means to achieve sustainable outcomes rather than an end in itself. China has initiated the establishment of WAICO, a coalition comprising 29 countries; while the EU's 27 member states enjoy privileged collaboration with several leading middle powers.

The EU and China should make deeper AI cooperation a goal — and pursue it with common agendas. Now is the moment to reckon and restart. With quantum AI and a new generation of powerful agents and robots on the horizon, the world cannot afford to wait for alignment to arrive on its own.

The author is the director of the Center for European Policy Studies in Brussels, Belgium. The author contributed this article to China Watch, a think tank powered by China Daily.

The views do not necessarily reflect those of China Daily.

Contact the editor at editor@chinawatch.cn.