The ITU’s AI for Good Global Summit in Geneva has come and gone once again. This time it coincided with the first report by the Independent International Scientific Panel on Artificial Intelligence. One of its conclusions is difficult to dispute: AI will reshape economies, public services, scientific discovery and national security. Like electricity and the internet before it, the countries with the capacity to develop and deploy AI will increasingly dominate markets, set standards and shape geopolitical influence.

As governments begin looking beyond the Sustainable Development Goals, AI is shaping education, healthcare, productivity, public administration and economic growth. In many respects, the post-2030 development agenda will be an AI agenda.

Yet the global conversation remains largely focused on the capabilities of AI models, their risks and the safeguards required. Far less attention is paid to the political economy of AI: who owns the data centres, who builds the AI models, and who profits most.

Division and extraction

Just a handful of companies are developing the world’s most advanced foundation models. Three US firms – Amazon, Microsoft and Google – control roughly two-thirds of the global cloud market, while Taiwan’s TSMC manufactures over 90 per cent of the world’s most advanced semiconductors. Nvidia supplies the vast majority of high-end AI chips that power frontier systems. Most governments will spend the coming decade deploying and regulating technologies developed elsewhere, running on infrastructure they neither own nor influence. The result is not simply an AI divide. it is rather a widening gap in capability, development and power.

AI is also creating a new extractive economy. The raw materials are data, public research, human creativity, energy and computing power. They are sourced globally, processed through infrastructure controlled by relatively few actors and transformed into products whose value largely flows back to the platform owners. Countries risk supplying the inputs of the AI economy while importing the intelligence built from them.

That imbalance is unlikely to correct itself. As AI models become cheaper and more capable, demand for chips, cloud services, data centres, electricity and water will continue to grow. Greater access to AI applications does not necessarily reduce dependence on the infrastructure behind them, as cloud computing has demonstrated. It lowered the cost of digital services while concentrating critical infrastructure in the hands of a few providers. AI is likely to deepen that trend.

A voice for all

This is why we need to shift from AI for good to AI for all. AI for all means more than universal access to increasingly powerful technologies. Countries need the talent, research institutions, computing capacity and public infrastructure to develop AI on their own terms and share in the value it creates.

Global AI governance has an important role to play, but we should also be realistic about its limits. Frontier AI will continue to be driven largely by companies with extraordinary financial resources and access to computing power. The United Nations will not determine their research agendas, nor should it pretend to. Its comparative advantage lies elsewhere. It can build international norms, promote inclusive governance and ensure that countries not developing frontier AI have a meaningful voice in shaping the principles under which these technologies are deployed.

International Geneva’s role

Geneva has a particular responsibility. For decades it has been where governments negotiate the rules for trade, telecommunications, health and humanitarian action, and artificial intelligence deserves the same institutional ambition.

The global AI summit process has evolved remarkably in just four years, since it began in the United Kingdom with the Bletchley Park AI Safety Summit in 2023, where 28 countries and the European Union met to address the risks of frontier AI. In Seoul, the process focused on a smaller coalition committed to advancing governance. Paris broadened participation again, with 58 countries, the EU and the African Union Commission endorsing its declaration. This year, New Delhi marked another important step, with 92 countries, alongside the EU and the International Fund for Agricultural Development, supporting the New Delhi Declaration.

When the summit comes to Geneva in 2027, the ambition should be greater still: to make it a truly global summit, bringing together as many states as possible to shape not only how AI is governed, but also how its benefits are shared. Geneva offers a unique advantage as most UN member states already maintain a diplomatic presence in the city. It lowers the barriers to participation, making it easier for governments from all regions, particularly developing countries, to take part in shaping the future of AI governance.

Switzerland should seize that opportunity by ensuring that the 2027 summit becomes the broadest global forum on AI to date. If artificial intelligence is becoming foundational infrastructure for the global economy, its governance cannot remain the preserve of a handful of governments and technology companies. More countries need the capacity not only to use AI, but to build it, shape it and benefit from it. AI for good inspired an important global conversation. AI for all should become its next ambition.

Martin Wählisch is an associate professor of international relations at the University of Birmingham and a member of its Centre for AI in Government. His research focuses on international affairs, AI governance and multilateral cooperation.

Read more on Science & Tech  Artificial intelligence  Opinion

.

.