India, Feb. 22 -- The world risks sleepwalking into an "AI divide" as stark as any inequality the internet age produced, with Africa home to fewer than a thousand GPUs while a single Chinese university may possess more, the United Nations' top official on emerging technologies has warned. Under Secretary General and Envoy for Emerging and Digital Technologies, Amandeep Gill, sat down with Anirudh Suri to discuss the 'looming AI divide,' the high-stakes risks of concentration of power and the human impact on children's mental health and cognitive development. Below are edited excerpts from the podcast, AI Futures: The Road to India AI Summit 2026, hosted by Suri, in collaboration with IndiaAI Mission and the ministry of electronics and IT (MeitY). There have been tremendous historic developments in the global governance of AI. UN member states came together in August 2025 to approve two new standing mechanisms for the global governance of AI: an Independent International Scientific Panel of 40 experts on AI, and a new Global Dialogue on AI Governance where all 193 member states will have a seat at the table. The member states had also asked the Secretary-General for options to bridge the AI divide. 2.7 billion people are still not on the internet and we have a looming AI divide; Africa has access to less than a thousand GPUs, which a single university in China may have. There is fragmentation of the international effort on AI governance. Therefore it's become more challenging for us to navigate the cross border impact of AI. This is the biggest concern that we have at the United Nations: the concentration of economic and tech power and the hubris that comes with it, which sometimes makes you do things that are risky. With AI, there is greater risk of that. We cannot afford to let that happen because this great divergence would occur between but also within the countries. Even in high-income countries we may have great disruptions in white collar jobs. Political leaders and civil servants need to take this much more seriously. Two, there should be a massive focus on talent development. If we can bring AI literacy into the general population, people will understand both the power and the limitations of this technology. And finally, there's the imperative of international cooperation. The leading AI nations need to realize that if you help other countries develop absorptive capacity, the size of the market will be much bigger than if most countries are just bystanders in the AI race. Absolutely. This is a very important risk. AI is compute-intensive, especially with the current emphasis on large models that are very energy hungry. I don't know if that's a wise choice. You have SLMs - small language models - that are sustainable and can do most of the jobs. We need to shift the emphasis from a competitive development of models, we need more collaborative development. As we build these data centers, these AI super factories, we can use this opportunity to also double down on renewable resources. It has to be both. In the Global South, the private sector may not have the capacity to buy compute or build data centres on its own. In the Secretary General's report on Innovative Financing Options for AI Capacity Building, we analyzed different countries. Certain countries are not going to have access to market funding, so the government will have to step in. The initial demand for private sector applications may also need to come from the government. I see this in Southeast Asia, Africa, and in places like India. On infrastructure, for example, under the National AI Mission in India, the government is taking the lead to subsidize access to 40,000 GPUs for researchers, developers, and smaller companies. In Vietnam and other countries in ASEAN, you see the development of context-specific, language-specific agentic AI which sits on edge devices. In Africa, a company is bringing 13,000 GPUs to the five biggest data markets in Africa. An exciting company from Nigeria is building a voice first AI approach rather than text-first LLM approach with tonal languages like Yoruba. AI innovation. I think adoption and diffusion will happen. But if you don't pay attention to AI innovation, then diffusion would be kind of self-limiting. My sense is that in the long run, AI will create more productive work opportunities though it's hard to predict numbers. In some sectors, the numbers are definitely going to go down. So you may not need as many data analysts or coders at the entry level. But then coders would use AI tools and be more productive and upskill. So whatever you're doing, if there's a way to bring AI into that, don't hesitate to experiment, do it. I'm not going to pick. At the end of the day, AI is not going to disrupt passion. It is not going to take away what makes your heart sing. So if it's poetry, go with poetry. If it's coding, go with coding. Sometimes it's not easy to find what your heart desires. You may have to experiment, you know, continue turning the radio dial till you hear the music....