AI Power Play, No Referees

20 Jan 2026 16:30h - 17:15h

Session at a glance

Summary

This World Economic Forum panel discussion, moderated by Ian Bremmer, examined how artificial intelligence is reshaping global power dynamics and economic relationships. The distinguished panel included IMF Managing Director Kristalina Georgieva, Microsoft President Brad Smith, Saudi Investment Minister Khalid Al-Falih, and Indian IT Minister Ashwini Vaishnaw.

Georgieva presented the IMF’s AI preparedness index, categorizing countries into three groups: those who make AI happen, those who watch it happen, and those wondering what’s happening. She highlighted that AI affects 40% of jobs globally, with advanced economies seeing 60% impact compared to 25% in low-income countries, creating an “accordion of opportunities” that widens inequality both between and within nations. The discussion revealed competing perspectives on global AI leadership, with both India and Saudi Arabia asserting their positions in the top tier rather than as followers.

Smith emphasized the importance of transatlantic cooperation in AI development while noting significant regional differences in data center deployment approaches. He warned about the growing digital divide between the global north and south, where AI adoption rates are 25% versus 14% respectively. The panelists discussed practical AI applications, from India’s public-private GPU sharing model training 10 million people in AI skills, to Saudi Arabia’s healthcare diagnostics reducing diabetic eye disease diagnosis time by 80%.

Key governance challenges emerged around weaponization concerns, the need for techno-legal regulatory approaches, and addressing entry-level job displacement for young workers. The discussion concluded with calls for increased investment in skills training, infrastructure development particularly in Africa, and international cooperation to prevent AI from exacerbating global inequalities.

Keypoints

Major Discussion Points:

– Global AI Power Distribution and Geopolitics: The panel discussed how AI is reshaping global power dynamics, with countries falling into three categories: those who “make it happen” (US, China), those who “watch it happen,” and those asking “what’s happening?” The conversation explored how AI capabilities are influencing international relationships and whether countries can chart independent courses versus aligning with major AI powers.

– Economic Impact and Labor Market Transformation: Significant focus on AI’s dramatic effects on employment, with the IMF calculating that 40% of jobs globally are “touched” by AI (60% in advanced economies). The discussion highlighted how AI creates a “tsunami” in labor markets, particularly affecting middle-skill jobs and young people seeking entry-level positions, while enhancing high-skill roles.

– Infrastructure and Investment Challenges: The panel addressed the massive infrastructure requirements for AI deployment, including data centers, energy needs, and compute resources. They discussed community resistance to data centers in the US, the need for public-private partnerships (as demonstrated by India’s GPU sharing model), and the importance of making AI infrastructure globally accessible.

– Digital Divide and Global Equity: A major concern about AI exacerbating global inequalities, with only 25% of the global north versus 14% of the global south currently using generative AI. The discussion emphasized the need for international cooperation to prevent AI from widening the gap between developed and developing nations, particularly in Africa.

– Governance, Regulation, and Sovereignty: The panel explored the need for “techno-legal” approaches to AI regulation, addressing issues of bias, data sovereignty, and the weaponization of AI platforms. They discussed the importance of countries developing their own AI capabilities and governance frameworks while maintaining international cooperation.

Overall Purpose:

The discussion aimed to examine AI’s transformative impact on global economics, geopolitics, and society from multiple national and institutional perspectives. The panel sought to address both the opportunities AI presents for economic growth and productivity, as well as the significant challenges it poses for employment, international relations, and global equity.

Overall Tone:

The tone was professional and analytical, with participants demonstrating both optimism about AI’s potential and serious concern about its disruptive effects. The conversation maintained a collaborative spirit despite some competitive undertones, particularly when panelists defended their countries’ AI capabilities and positioning. The tone became more urgent when discussing the need for proactive governance and the risks of leaving developing nations behind, ending on a somewhat cautionary but constructive note about the need for immediate action on skills development and global connectivity.

Speakers

Speakers from the provided list:

– Ian Bremmer – President of Eurasia Group, Panel Moderator

– Kristalina Georgieva – Managing Director of the International Monetary Fund (IMF)

– Brad Smith – President and Vice Chair of Microsoft

– Khalid Al-Falih – Minister of Investment of the Kingdom of Saudi Arabia

– Ashwini Vaishnaw – Minister for Economic Electronics and Information Technology of India

– Audience – Various audience members who asked questions during the Q&A session

Additional speakers:

– Pranjal Sharma – Columnist from India focusing on the intersection of tech and policy

– Ishan Pratap Singh – 22-year-old entrepreneur from the Global Shapers Community, New Delhi hub

– Maria – AI developer from Portugal working on drug discovery, also a Global Shaper

Full session report

AI and Global Power Dynamics: A Comprehensive Analysis of Economic Transformation and Geopolitical Implications

Executive Summary

This World Economic Forum panel discussion, moderated by Ian Bremmer, brought together leading voices from international finance, technology, and government to examine how artificial intelligence is fundamentally reshaping global power dynamics and economic relationships. The distinguished panel featured IMF Managing Director Kristalina Georgieva, Microsoft President Brad Smith, Saudi Investment Minister Khalid Al-Falih, and Indian IT Minister Ashwini Vaishnaw.

The discussion revealed a complex landscape where AI’s transformative potential is matched by significant challenges in ensuring equitable global development. Georgieva’s presentation of the IMF’s AI preparedness index established a framework categorising countries into three distinct groups: those who “make AI happen,” those who “watch it happen,” and those left “wondering what’s happening.” This taxonomy became the organising principle for examining how AI affects 40% of jobs globally, with advanced economies experiencing 60% impact compared to 25% in low-income countries, creating what Georgieva termed an “accordion of opportunities” that risks widening inequality both between and within nations.

The conversation challenged conventional assumptions about AI leadership, with both India and Saudi Arabia asserting their positions as first-tier AI powers rather than followers. This pushback against traditional categorisations highlighted the evolving nature of global AI competition and the importance of comprehensive national strategies spanning infrastructure, skills development, and governance frameworks.

Global AI Power Distribution and the Three-Tier Framework

Challenging Traditional Classifications

Kristalina Georgieva’s opening remarks established a framework for understanding global AI readiness: “We created the Index of Preparedness to see how well are different countries ready for AI. I would put them in three categories. Those who make it happen, those who watch it happen, and those who are saying, what the heck is happening?” The index evaluates countries across four dimensions: physical infrastructure, labor skills, diffusion capabilities, and regulation and ethics frameworks.

This categorisation immediately sparked debate, with both Ashwini Vaishnaw and Khalid Al-Falih challenging their countries’ perceived placement. Vaishnaw argued forcefully that India should be classified in the first tier, stating: “India operates across all five layers of AI architecture… application layer, model layer, chip layer, infrastructure layer, and energy layer.” He cited Stanford rankings and comprehensive capabilities to support India’s first-tier status.

Similarly, Al-Falih positioned Saudi Arabia as a first-tier player, emphasising the kingdom’s comprehensive AI infrastructure including data centres, sovereign data policies, and the development of the first Arabic large language model. This pushback revealed fundamental disagreements about what constitutes AI leadership and how countries should be evaluated.

Reframing AI Competition

The discussion revealed competing visions of AI competitiveness. Vaishnaw challenged prevailing assumptions about the geopolitical value of large AI models, arguing that “ROI doesn’t come from creating a very large model. 95% of the work can happen with models which are 20 billion or 50 billion parameters.” This perspective reframed the entire discussion about AI leadership, shifting focus from raw computational power to practical deployment and return on investment.

Al-Falih supported this efficiency-focused view, suggesting that AI will become “a global commodity like hydrocarbon energy,” requiring optionality and investment across multiple countries and technologies. Both emerging economy representatives advocated for practical accessibility over the pursuit of the largest possible models.

Brad Smith emphasised the importance of transatlantic cooperation, noting that “70% of foreign investment into the US comes from NATO countries, 40% of American exports go to NATO countries,” whilst acknowledging the broader US-China competition on innovation leadership and global technology diffusion.

Economic Impact and Labour Market Transformation

The Employment Challenge

Georgieva characterised AI’s impact on employment as creating significant disruption across the global workforce. Her analysis revealed that “40% of jobs are touched by AI, either enhanced or scrapped or changed quite significantly,” with the impact varying dramatically by economic development level—60% in advanced economies compared to 25% in low-income countries.

The IMF’s own experience provided a concrete example of this transformation. Georgieva noted that in “translation and interpretation went from 200 to 50 people, 150 replaced by AI,” while research analysts were enhanced rather than replaced. This pattern of AI replacing some roles while augmenting others was reflected in broader labour market trends, with Georgieva noting that “in the United States, one in 10 job advertisements would have a new skill that didn’t exist before.”

The discussion highlighted particular concerns about the “squeeze” on middle-skill workers and young people entering the job market. As Georgieva explained, this creates an “accordion of opportunities” where AI simultaneously enhances high-skill roles whilst potentially displacing positions that traditionally served as entry points for young workers.

Skills Development as National Priority

Despite disagreements on other issues, all panellists converged on the critical importance of massive investment in skills development and education. Vaishnaw outlined India’s ambitious programme to train 10 million people in AI skills and provide subsidised access to compute resources through public-private partnerships.

Smith argued that “investment in skilling is more feasible for governments than data centre investment and can stimulate private sector demand.” However, approaches varied significantly. India’s model emphasised direct government provision of AI infrastructure through a common compute facility with 38,000 GPUs available to the entire population at one-third the cost of other countries, contrasting with Smith’s preference for government focus on skills development whilst leaving infrastructure primarily to private sector investment.

Measuring Economic Impact

The discussion highlighted significant challenges in quantifying AI’s economic benefits. Georgieva acknowledged that measuring AI’s impact on global growth remains difficult, with estimates varying widely. Al-Falih provided concrete examples of AI applications in Saudi Arabia, including reducing “diabetic eye disease diagnosis time by 80%” and achieving a “30% reduction in chemical usage in desalination,” along with implementing wind turbine bird detection systems. These specific applications demonstrated measurable benefits in targeted sectors whilst highlighting the challenge of aggregating gains into broader economic metrics.

Infrastructure Challenges and the Digital Divide

The Widening Global Gap

One of the discussion’s most concerning revelations was the growing disparity in AI adoption between developed and developing regions. Smith presented statistics showing that “generative AI is used by 25% of the global north population but only 14% in the global south, with the gap widening.” This digital divide threatens to entrench existing global inequalities and create permanent disadvantages for developing nations.

Georgieva highlighted the scale of infrastructure investment needed, specifically calling for “$15 billion to connect every African citizen to the internet.” This figure underscored both the magnitude of the challenge and the concrete steps needed to address global connectivity gaps.

Community Resistance and Local Concerns

Unexpectedly, the discussion revealed that immediate political challenges around AI adoption focus more on infrastructure concerns than employment displacement. Smith identified community resistance as a significant obstacle, noting that “local communities blocked $98 billion of private sector investment in just the third quarter of last year,” primarily due to concerns about electricity costs, water usage, and limited local benefits.

This insight suggested that successful AI deployment requires addressing local community needs and ensuring tangible benefits for areas hosting AI infrastructure. Smith emphasised the need for tech companies to provide community assurances about electricity rates, water usage, and local job opportunities to gain social licence for data centre projects.

Innovative Infrastructure Models

India’s approach to AI infrastructure emerged as a potential model for other developing nations. The creation of a common compute facility with 38,000 GPUs available to the entire population at significantly reduced costs demonstrated how public-private partnerships could democratise access to AI resources whilst maintaining cost-effectiveness.

This model addressed both affordability challenges and the need for broad access, potentially offering a pathway for other emerging economies to participate meaningfully in the AI economy without requiring massive individual investments in infrastructure.

Governance and International Cooperation

Technical Solutions to Regulatory Challenges

The discussion revealed sophisticated thinking about AI governance that moved beyond traditional regulatory frameworks. Vaishnaw advocated for what he termed “a techno-legal approach to regulation,” arguing that effective AI governance requires creating technical tools to counter harmful effects like bias rather than relying solely on laws.

This approach recognised that AI’s complexity demands technical solutions integrated with legal frameworks rather than purely legislative responses. The perspective was supported by Al-Falih’s emphasis on sovereign AI models and Smith’s focus on international agreements to prevent technology weaponisation.

Security and Weaponisation Concerns

Smith’s observation that “every technology is both a tool and a weapon” provided a framework for understanding AI’s dual nature. He warned about the potential for AI to be weaponised and highlighted the need for international dialogue to prevent technology from becoming a geopolitical weapon.

The discussion touched on concerns about digital platforms being weaponised and the need for international cooperation to address these challenges whilst maintaining innovation and competitive dynamics.

Cultural and Linguistic Considerations

Al-Falih’s emphasis on developing Arabic language models highlighted important cultural dimensions of AI governance. The need for AI systems to reflect local values and contexts suggested that effective governance must address not only technical and economic concerns but also cultural relevance and linguistic diversity.

Key Areas of Agreement and Persistent Challenges

Fundamental Consensus

Despite disagreements on specific approaches, the discussion revealed consensus on several fundamental principles. All speakers agreed that AI represents a transformative technology requiring comprehensive national strategies and that skills development and education are essential responses to AI’s labour market impact.

There was also shared concern about the growing digital divide between developed and developing regions and the need for targeted interventions to prevent further inequality.

Ongoing Disagreements

Significant disagreements emerged around implementation approaches and country classifications. The most notable dispute centred on how to measure and classify AI readiness, with both India and Saudi Arabia challenging the IMF’s framework and asserting their positions as first-tier AI powers.

Important differences also emerged regarding the economic viability of large AI models versus smaller parameter models, and the appropriate balance between government and private sector roles in AI infrastructure development.

Conclusion

This comprehensive discussion revealed AI’s profound implications for global power dynamics, economic development, and social equity. The conversation moved beyond simplistic narratives about AI competition to explore nuanced questions about measurement, governance, and inclusive development.

The emergence of alternative models for AI development, particularly India’s public-private partnership approach and the emphasis on cost-effective smaller models, suggests that the future of AI may be more distributed and diverse than current competition narratives suggest. The strong pushback from emerging economy representatives against traditional categorisations indicates a more multipolar AI landscape may be developing.

However, the discussion also highlighted significant risks, particularly the widening digital divide and the potential for AI to exacerbate existing global inequalities. The emphasis on community resistance to AI infrastructure and the need for local benefits suggests that successful AI deployment requires addressing social and political dimensions beyond technical capabilities.

The convergence on skills development as a critical priority, combined with innovative approaches to infrastructure sharing and governance, provides a foundation for international cooperation. However, persistent disagreements about measurement frameworks and implementation approaches suggest that coordinated global responses to AI challenges will require continued dialogue and compromise to address technical, economic, social, and political dimensions simultaneously.

Session transcript

Ian Bremmer

And hello, everybody. I am Ian Bremmer, president of Eurasia Group, and this is AI Power Play. I did not come up with the name for this panel.

You’ll be very pleased to hear. We have a lot to talk about. I’m going to very briefly introduce the very distinguished panel of guests.

Immediately to my left, the managing director of the International Monetary Fund, Kristalina Georgieva, Brad Smith, the president and vice chair of Microsoft, Khalid Al-Falih, the minister of investment of the Kingdom of Saudi Arabia, and last but certainly not least, Ashwini Vaishnaw, the minister for economic electronics and information technology of India.

So a pretty good global breakdown, interests, responsibilities. We’re going to try to get to a couple of questions at the end. If you have something really smart to say as we get towards the end, try to make eye contact.

We’ll see what happens. So I want to start off with Kristalina. And we’re going to begin with sort of the geography, the geopolitics of what’s happening in AI right now.

I mean, you’d think it’s all about Greenland and Davos this year. But actually, there’s another Davos going on, which is very fast moving and involves an incredible amount of investment. It’s one of the reasons why you just upgraded the economic outlook.

Talk about, from your global perspective, how is power changing around the world with AI?

Kristalina Georgieva

Oh, definitely it is. When we look around the world and we see what is injecting resilience and strength in the world economy, AI definitely is a factor, and so is the advancement of private activity over the last decades. The state moving out of the economy, letting private business do its part.

And those two things work extremely well. So we created the Index of Preparedness to see how well are different countries ready for AI. I would put them in three categories.

Those who make it happen, those who watch it happen, and those who are saying, what the heck is happening? And actually, on a serious note, we look at four things. We look at physical infrastructure, we look at labor skills capabilities, we look at diffusion, AI may be happening, but if it is not changing the economy, it’s not so relevant, and then we look at regulation and ethics.

And then we rank countries. Not surprisingly, we have at the top a very small group of countries. Now, in the current environment, probably would be interesting to recognize that the United States, Denmark, and Singapore are the top three.

And of course, China is with a very potent capabilities, but China is so big. that you can’t get that high in this ranking. Not surprisingly, we have many emerging market economies with comparative strength.

Saudi Arabia would be one of them. Actually, India would be one of them because of the bet on IT that India is making that are in the higher spectrum. But low-income countries and a very big number of mid-size, middle-income countries are way, way behind.

And they’re behind on all four. So the world as a whole is already experiencing the arrival of AI. But I do worry about the accordion of opportunities that are much more present in some places than in others.

And then we look at the impact on labor market. It is quite dramatic. We calculate that on average, 40% of jobs are touched by AI, either enhanced or scrapped or changed quite significantly.

Without implications for better pay. In advanced economies, this is 60%. In low-income countries, 25, 26%.

Again, accordion of impact. Now, my main message here is the following. This is a tsunami hitting the labor market.

And even in the best-prepared countries, I don’t think we are prepared enough. And if you allow me to add one point.

Ian Bremmer

It’s just gonna take away from your time later, Kristalina.

Kristalina Georgieva

Oh, I know, I know. I know you. You are like an accountant.

You help me, you help us to account. And I’ll give you, you know, two minutes from the next one.

Ian Bremmer

Okay, okay.

Kristalina Georgieva

So what I want to say is that we did something very interesting. We said, okay, how much are actually jobs changing, really? And what we found out that in advanced economics, in the United States, for example, you read advertisements and one in 10 would have a new skill that didn’t exist before, at least one new skill, maybe two, maybe three, maybe four.

And that these guys would get better pay. So what is the impact? The impact is they have more money in their pockets, they go to restaurants more, to movies more, so demand for low-skill labor in services goes up.

And in fact, who is squeezed? These poor guys in the middle, in the jobs that are not enhanced or are imposed more. And the other category of people really, really affected young people who go for their first job.

Ian Bremmer

So. Power is changing dramatically.

Kristalina Georgieva

I owe you one minute and 26 seconds.

Ian Bremmer

Future panels, I’m taking this off as well. So many, many issues of both power distribution inside countries, which I hope we’ll get to, but first around the world. And those three categories that you remind everybody at the beginning, you make it, you watch it being made, or you have no idea what’s being made.

The countries in those categories don’t necessarily always align very well. First category, we’ve got the US and China, very little cooperation, virtually none on this issue, arguably decoupling. Second group, a lot of movement.

And as we’re seeing this week, a lot of concerns about the nature of like the transatlantic relationship, how that plays out on AI and technology, very important. Brad, talk us through the implications of this.

Brad Smith

Well, certainly you see the US and China competing on two very important levels. One is the leading edge of innovation. who is making the best chips, who is producing the best models.

But the second aspect of competition is really diffusion or adoption. As each country and the companies from each country are working to spread their technology around the world. I think it’s easy to forget in a more global economy just how important the transatlantic ties are.

When you have a conversation on a week like this about an island like Greenland, of course it’s a national security topic, but it needs to be considered, I think, in the context of these economic ties.

When it comes to AI, I think people in Europe often look at the United States and say, well, we need this or we need that, but they forget they have ASML, they have Ericsson, they have Nokia, they have SAP, they have Siemens.

There is a very important tech sector in Europe already. And if you look at the ties across the Atlantic, it’s easy for Americans to forget that 70% of foreign investment into the United States comes from the rest of NATO. 40% of American exports globally go to the rest of NATO.

So we always need to remember the context in which these issues need to be considered.

Ian Bremmer

If you think about the changes that AI is having on the geopolitical environment, Brad, would you say that AI is doing more to bring countries together, is doing more to force countries apart, or is it, on average, kind of absent from that conversation?

Brad Smith

I would probably say, on average, it remains absent, but it is going to become more important every year. And when you look at the kingdom of Saudi Arabia, when you look at India, I think you see the future and you see the role that AI is playing there already, and you see the role that that will play, I think, in both growing their economies, but connecting their economies in new ways with the rest of the world.

Ian Bremmer

So Ashwini, incredible talent base, fastest-growing major economy in the world. an AI player, but clearly in the second grouping that Kristalina’s talking about, have to think about alignments with the United States or China to a degree. But how much do you have to think about that as a country like India?

How much can you chart your own course irrespective of what the Americans and Chinese are doing?

Ashwini Vaishnaw

Actually, clearly in the first group, and the reason for that is there are five layers in the AI architecture. The application layer, the model layer, the chip layer, the infra layer, and the energy layer. We are working on all the five layers, making very good progress in all the five layers.

On the application layer, we will probably be the biggest supplier of services to the world. Go to an enterprise, understand the business of enterprise, understand the working of that enterprise, and provide that service using AI applications. That’s going to be the biggest factor of success or successful deployment of AI because that’s where ROI comes from.

ROI doesn’t come from creating a very large model. 95% of the work can happen with models which are 20 billion or 50 billion parameters. We are creating a bouquet of such models, we already have.

We already have a bouquet of such models which are now being deployed in multiple sectors to increase the productivity, to increase the efficiency, to increase the effective use of technology. So our focus is very much on making sure that AI diffusion happens in a very big way. And I don’t know what the IMF criteria has been, but Stanford places India as third in terms of AI penetration, in terms of AI preparedness, and in terms of AI talent.

All the three, actually on AI talent, it is number two. So I don’t think your classification in the second bouquet is right. It’s actually in the first.

Ian Bremmer

So talk about India. in an environment where the AI shifts in geopolitics are happening so quickly. Again, this was a conversation that didn’t even exist at the World Economic Forum five years ago.

And you now have radically different capabilities in different countries around the world. India has become a geopolitical player in its own right in the last 10 years. Didn’t even want to play geopolitics before Modi came in.

How does AI play into that capacity around the world? And to what India wants to be on the global stage?

Ashwini Vaishnaw

Listen, if you look at the way AI is shaping geopolitics, what can a country which has a very large model, let’s say, without taking any names, can it switch off that model? Fair, switch it off. What will happen to a country like India?

We have our own bouquet of models, which can be used for 95% of the work that we require to do. So, does creating a large model give you a geopolitical power? I don’t think so.

It might actually be causing certain conditions where the people who are creating those large models might go bust in the coming years. You never know. They might go bankrupt in the coming years.

The situation is, we have to understand the economics of this new, I call it fifth industrial revolution. We must see the economics of the fifth industrial revolution. The economics of this fifth industrial revolution is going to come from ROI.

ROI is going to come from deploying the lowest cost solution to get the highest possible return. If you have a 50 billion parameter model, you can deploy it using one GPU. And if you have a 30 billion parameter model, which is absolutely good for 80% of your work, you don’t even require a GPU.

Where is the geopolitical thing in that? You have a large number of CPUs working in the entire world and people are coming with custom silicon. from amazing number of companies and countries where you wouldn’t require to be dependent on any particular country.

So that the so-called geopolitical edge that you are probably hinting at is not there.

Ian Bremmer

So given that, I mean, without naming names, Khalid, Ashwini was just talking about, you know, a lot of people that have been raising a lot of money in the United States with models that he thinks aren’t gonna be as effective as India’s model.

Whatever bucket they’re in. Now, you’re the one that’s thinking about deploying enormous amounts of assets for your country, both in Saudi Arabia and all over the world. And as you’re thinking about the most revolutionary set of technologies out there, what are you looking for that’s gonna make you convinced that this is a winner, this is a loser, as these different countries and different companies are doing very different things?

Khalid Al-Falih

Yeah, well, first, let’s, I think everybody agrees that AI as a general purpose technology is truly the transformation of this century and of our time. However, like transformation before, I think it will quickly, quickly be commoditized. And it’s not going to be held in any one company or any one country.

The race is on. Everybody wants to build the infrastructure for it, but it is, I think the essence of AI’s power is it has to be accessible. So in other words, diffusion is not just within economies that have to compete, but I believe it has to be done globally.

From our perspective, as an energy producing nation, the diffusion of hydrocarbon energy that Saudi Arabia produced for nine decades has enabled tremendous, tremendous human development in countries like India, China, United States, Europe, where that energy transformed.

economies, and people’s lives, the Internet did the same thing. There were a few companies, Microsoft was one of them, that had a lead, but ultimately it’s coming from across the world, including emerging economies and developing nations. We are manufacturing servers for data centers today in Saudi Arabia through partnerships.

So we believe, we believe that optionality is very important. We don’t know. We don’t know who’s going to be ahead four or five years from now.

I think it’s clear U.S. companies that have broke out of this technology race are ahead today. But our, the essence of Saudi Arabia’s strategy towards AI, it’s a huge boost for diversification.

Option 2030 launched by His Royal Highness the Crown Prince almost 10 years ago was underpinned by diversification using new economies, new economies. So the energy transition, although it may appear to be competing with our traditional resource of hydrocarbon energy, is actually a pillar of our diversification. We’re going to be 50% renewables in our electricity mix, and that renewable will power, will power AI that is going to be a global commodity that we will, that we will work with.

And just like India, we know this is not just about infrastructure, data center, and the energy competitive advantage that we believe Saudi Arabia is second to none. We’re investing across the technology stack and applications and LLMs and in connectivity. Again, because we believe that this is going to be a global good.

So just as important as building the data hub that Saudi Arabia is building, we need to be connected. And we are connected to Europe, Asia, countries like India, China, Japan, because we want that data, that AI power. to be transmitted across borders and across economies.

Ian Bremmer

Now, as we think about AI as having dual use capabilities, as the United States government is leaning more into industrial policy with other countries around the world as the Chinese also have a lot of leverage in the investments that they make and that they accept, are you feeling any geopolitical pressure in the decisions that you make from an investment perspective into these areas?

Khalid Al-Falih

Quite frankly, we have invested across the globe and we have invested in Saudi Arabia with companies from East and West. U.S. is huge today because of their technology lead, as I mentioned, but we’ve also had investments with Chinese companies in China and digital technologies.

We’ve invested with Japanese companies, Korean companies across the technology sectors that we invest in, so optionality is very important. It’s something we have now and we protect because we believe that we are the owners of our own destiny and we will not let go of that.

Ian Bremmer

So if that’s the broad geopolitics that we’re looking at, Kristalina, you brought up at the beginning also that inside countries, lots of people that are not necessarily benefiting. You’re already seeing improvements in economic numbers from productivity and AI. You’re already seeing that.

Kristalina Georgieva

What we are seeing is that there are indeed jobs that get enhanced and we also see jobs that are replaced by AI. This is happening categorically. It is hard to have a consolidated and broadly accepted definition of how we measure, the way we measure at the IMF we have a fairly big range of impact on growth, global growth, from 0.1 to 0.8 percent.

Let me just say that 0.8 percent is huge. If we get 0.8 percent boost on productivity, this would make global growth now higher than in the pre-pandemic period. 0.1 is kind of modest.

And then the question is how do we know where we land? And what we are doing is we are working with other researchers to try to capture productivity enhancement. And I can tell you it is not easy because, is it because of AI or is it because organizationally there are improvements that are made.

I’ll tell you in my organization the most visible places where AI is boosting productivity are two. One, translation and interpretation. We got from 200 to 50.

150 are replaced by AI. And in research analysts, there we are actually just able to do much more high-quality research. In other words, it is enhancement not a replacement.

My big worry is that you go in communities where AI is not present. And there, how do people prepare for it? There is nothing happening now.

And actually we did something, I’m not going to vouch 100 percent this is correct, but we did a review to see countries where there is demand for AI skills, whether this demand is matched with supply of AI skills.

And we found very interesting picture. In some countries, demand is very high, supply is low. Some actually, we are developed countries.

In some countries, supply is very high, you have tons of skilled people, but the economy is not absorbing that supply. And the sad story is the countries where there is neither demand nor supply. And that risk of this massive divergence is at the country level then mirrors within a country.

Within a country.

Ian Bremmer

So Brad, Kristalina, with this evocative accordion analogy, which is affecting not only countries but also populations, you’ve been on the front lines of this in the United States. You see now suddenly this raise in opposition to data centers and people saying, oh, we’re unhappy, we don’t think of the water, it’s electricity, we’re not gonna see the jobs. I mean, and a lot of people expect that given the shift in skills that are necessary, given the transition, that you’re gonna see that America’s ripe for another wave of populism.

How do you think the companies, not just Microsoft, but across the board, need to get ahead of this? What do they need to do that they’re not necessarily doing now?

Brad Smith

Well, the big issue of the United States is, as you’ve just mentioned, Ian, it really is around communities with data centers. And yeah, it’s coming at a time when the number one political issue, the number one economic issue is probably affordability, the impact of inflation. So people see these large construction projects, they know that those jobs actually are good jobs, skilled labor jobs, not just.

You have people doing construction work, but it’s skilled electricians, skilled pipe fitters. But they also know that once the construction is over, the number of jobs is still significant in the hundreds, but not higher. But what they’re really asking is, what does this mean for me and my family?

Are we gonna pay higher electricity prices? Is the water pressure in our showers gonna be impacted? Who’s really going to get these jobs?

Is it gonna be my family and my neighbors, or is it gonna be somebody who moves into town from somewhere else? Those are completely legitimate questions. And I think it’s incumbent on all of us in the industry in the United States to address them head on and offer the kinds of assurances that people need, that we will invest in electricity so that people’s electricity rates don’t increase, that we’ll replenish more water in these communities than we use, that we’ll work with the kind of training, whether it’s skilled labor or IT jobs or other ongoing jobs, so that local residents can fill them.

And I think as data centers spread, inevitably people in other parts of the world will ask the same question. If people benefit, will it be us? And I think people have a right to expect their share of the benefits.

Ian Bremmer

And to be fair, I wanna repeat this. You would say that today, a much bigger issue for the tech companies and AI is response to these concerns about data centers and affordability than concerns about job loss or job transferability in terms of right now.

Brad Smith

In terms of the politics of 2026, for sure. I mean, the world is sort of a quilt, if you will, with different colors of fabric when you look at the construction of data centers. In the US, local communities blocked $98 billion of private sector investment in just the third quarter of last year.

You come to Europe and governments want to spend taxpayer dollars to subsidize data center construction. You look at the Middle East, and there’s obviously a major economic strategy of creating an export, a token export or exports, if you will, out of, say, Saudi Arabia, the UAE, and the like. China probably has an ability to go faster in some ways when it comes to bringing electricity online and getting infrastructure built.

You go to Africa, wow, we just need, we’re still- There isn’t electricity. We need more electricity, and we need more data centers.

Ian Bremmer

So Ashwini, when you think about how AI can be used, technology, India already did an incredible job with the Aadhaar, suddenly you’re able to bring public services to people, and you’re improving growth, and you’re improving their per capita income without actually blowing out the budget.

What can AI do? How are you deploying AI to further that message? How fast are we going to see it?

Where are we going to see it?

Ashwini Vaishnaw

Absolutely. Every sphere of life and economy, we are focusing on diffusion of AI, and in a very systematic way. So, okay, what is the biggest constraint?

The biggest constraint is availability of GPUs. How do you solve that constraint? To solve that constraint, we decided to have a public-private partnership in which we empaneled 38,000 GPUs, which are a common compute facility available to the entire population.

Unlike in many rich countries where the big tech basically controls the access to GPU, we decided to create a public-private partnership in which the common compute facility is enabled by the government and subsidized by the government, so that entire population, all the students, researchers, start-ups, they all get access to it.

And the cost is practically one-third of the cost in most of the countries. That is first. Second, having a set of free…

a bouquet of models which basically meet most of your needs. Third, making sure that people understand this technology, learn this new technology. Already we are going through a program for 10 million people to be trained on AI skills.

Four, making sure that the IT industry, which is a very significant industry in our country, pivots in a very systematic way towards providing services to Indian companies as well as to the global companies using AI as a very effective enabler of efficiency, enabler, a multiplier of productivity in any operation that they do.

So that’s a very systematic way we are moving.

Ian Bremmer

Thank you. So, Khalid, I want to ask you to look inside the country. Last time I asked you to look outside on winners and losers.

The funny, the crazy thing about Saudi Arabia is that you’re kind of starting from scratch, right? I mean, 10 years ago, you’d go to Saudi Arabia and, I mean, diversification, they’d talk to you about Petrochem. And today, like, the economy is suddenly truly becoming diverse.

It’s healthcare and it’s tourism. 10 years ago, you barely had women in the workplace. And today, like, it’s an astonishing involvement in every part of society.

So, given that, like, where are the, AI is kind of hitting your economy almost from scratch as a diversified and integrated society with, what, 30 million people, whatever it is. How is, how are you deploying AI with people who are looking completely differently at the world today?

Khalid Al-Falih

Well, first of all, I would disagree with coming from scratch. You’ll be surprised how tech-savvy and tech-native the Saudi population is. 70% of the Saudi population are below the age of 35 and the majority of them are as comfortable with digital technologies as any other country.

in the world in terms of policy and regulation. When Vision 2030 was launched by His Royal Highness the Crown Prince, we identified technology as the key enabler, AI included, and started building the policy, the regulation, the platforms within the country, the data centers for the sovereign data sovereignty that was key for portions of the data.

But in addition to data privacy, we had an open data. We talked about diffusion and access to compute, but access to data to achieve the same purpose of research, drug discovery, productivity improvement, having a policy also of open access to data was a pillar that was launched before COVID.

This is back in 2018 and 19. We have a ministry level organization called SADAYA, the Saudi Arabian, the Saudi AI and data authority that regulates and sets the guidelines with flexibility to allow the evolution of AI to take place and the innovation to take place, but it also protects and regulates and make sure ethical standards.

Within the UNESCO, Saudi Arabia was one of the first country that proposed guidelines for the country on ethics in AI. Then we create a national champion, which is co-invested by the PIF or sovereign wealth fund and our own code that is investing at scale way beyond the need of Saudi Arabia and the Middle East. Because as I mentioned in my first question, we think AI is going to be a very key pillar of the global economy and it’s going to be globally traded and therefore connectivity.

And we are playing to our strengths, but we are also quickly. bridging the gaps that we recognize we have. So we’re developing our own LLM.

Saddaia developed the first Arabic large model called Alam, and now it’s owned by Humane, which is our national champion, and they’re building on it. We have applications. I can tell you many, many use cases where over the last few years we have deployed, we have deployed AI applications in healthcare.

It has cut the time to diagnose diabetic eye disease by 80%, and the rate of, and the thousands of man hours of clinical physician hours have been saved in that case. We have our renewable energy company, Aqua Power. They have cameras on the blades of their winds that identify birds approaching from far away, and if they think they’re coming near the windmill, they shut.

Ian Bremmer

You gotta tell President Trump about that.

Khalid Al-Falih

They shut. Okay, they, the same company, have reduced their chemicals and water desalination by 30% using AI, and the use cases are abundant. So we’re not starting from scratch, going back to Kristalina.

We’re in your first category.

Kristalina Georgieva

I have not declared either India or Saudi Arabia in the second category, so I stand to accept your claim.

Ashwini Vaishnaw

Thank you.

Ian Bremmer

Everyone is in the first category going forward. Participation prizes, there’s no problem.

Kristalina Georgieva

I’m not gonna go that far.

Ian Bremmer

You wanted to jump in.

Brad Smith

I think we should, let’s not talk about Saudi Arabia or India for a moment, but let’s just talk about the global north as a whole and the global south as a whole. You know, the diffusion report that we released a couple of weeks ago estimated that today, generative AI. is used by 25% of the population in the global north, but only 14% in the global south, and the gap is getting wider.

So if you zoom out and you say, what does this mean for the future of the world? I put it in the context of infrastructure and the history of infrastructure. Why is the world so divided economically?

Well, I’d say a large part of it is because during the entire colonial era, the colonial powers invested in building railroads in places like India and across Africa. They were important to move troops, to control territory, to extract minerals, but they never built power plants for the population. Will AI exacerbate this divide or will it close it?

It’s only gonna close it if we embark on building infrastructure across the global south, and we need strategies that stimulate demand, that furnish supply. India, Saudi Arabia, they’re on the right track. It’s hard to look at Africa as a whole and be equally optimistic, and we need to think about that as well.

Ian Bremmer

So as we move into the last segment, and if anyone, question or two, raise your hand. I will try, I see a couple. Okay, we’ll try to get there.

I wanna talk a little bit about the governance environment, and you just mentioned a big problem, which is absent international governance, trying to figure out how you’re gonna build this infrastructure.

I mean, in this environment, it’s hard to say that the Americans are more committed to that than they have been the last five, 10 years. China’s been pretty inward-looking in many ways the last five, 10 years. Those are some big challenges.

This space is moving a lot faster than governance capabilities. So if you could wave your magic wand over the next few years, with most of the leaders that we presently have now, what are one or two things that you think are feasible that would actually improve governance, that would make a positive difference in addressing both the challenge that you just raised, but also some of the others that we’ve been talking about the last 30 minutes?

Brad Smith

Well, the first thing I would say is look back at the skilling. topic. Because it’s a lot cheaper for governments to invest in skilling than it is, frankly, to invest in data centers.

But if you invest in skilling, you start to stimulate demand for the private sector to invest in data centers. You can skill government employees. That has proven to be a very effective strategy for kick-starting demand.

You can, you know, skill data scientists. You can do that through universities. That’s where you start to get people who can then build the local applications.

And I just think it is a critical element that is not getting the attention it deserves.

Ian Bremmer

Khalid, what is the piece of governance or regulation that if you saw it in a country around AI, you would say, this will make me materially more interested in investing in it? In governance? Yes.

Khalid Al-Falih

Well, you know, generally I think the concern we have with AI is biases in the data sets that the models have been learning from. So, as someone who’s coming from Saudi Arabia and the Middle East, it’s important for us to also build our own sovereign models and to make sure that unbiased data is fed and taught to the models so that they are fit for purpose for us.

They need to be consistent with our values. And our value sets have to be protected because AI, as we gain all of the efficiency and productivity and competitiveness that we inevitably will gain, we also don’t want to lose our value and our character.

And that, I think, is an issue of global governance where we have to recognize that these models and these applications cannot be developed in one country, say Europe, US, China, whatever it is, and then assume to be relevant and applicable to every other country.

Ian Bremmer

OK, last comment before I go to a couple of questions. Ashwini, I wanted to say, I know that you and the way you think about regulation and governance in India are not just thinking about the laws that you’re creating, but also how you use the technology to actually create and implement.

I’m wondering if you could share just a moment on that with the audience.

Ashwini Vaishnaw

Absolutely right. In case of technologies like AI, it’s very important to have a techno-legal approach to regulation. It cannot be just a law that you pass and believe that everything will fall in place.

You have to create technical tools and technologies to counter the harmful effects of, for example, bias as Khalid said. For example, detecting defects with accuracy which can be taken to a court and be properly judicially checked. So those kind of things you require techno-legal approach.

We are following techno-legal approach, creating solutions which mitigate bias, which help in detecting defects, which make sure that unlearning can be done properly before you deploy a model in an enterprise.

So those kind of technologies we are developing.

Ian Bremmer

Thank you. So I’m going to try to take one question to begin. Please recognize yourself.

If you can be quick, I’ll try to get to two. We’ll see what we get.

Audience

So I’ve been asked to stand up, so I shall. I’m Pranjal Sharma. I’m a columnist from India.

And I focus on the intersection of tech and policy. The key word to use is weaponization. We’ve seen weaponization of digital platforms and AI.

We saw a case in India where a company’s digital footprint was switched off within seconds. We saw two justices of the International Court of Justice face the same. So my question to all of you is, how critical is this threat of weaponization of digital platforms, including AI?

And is that therefore driving this piece of tech and AI sovereignty beyond just the data sets and controlling data?

Ian Bremmer

Brad, that feels like it’s most for you.

Brad Smith

Well, you know, every technology is both a tool and a weapon. There’s a book with that title. And I think the, unfortunately, just as there’s almost an infinite way, number of ways to turn AI into a tool, it can also be turned into a weapon.

I would highlight two issues that I think are really paramount today. One is the use of AI to increase cyber threats. And that just is a classic case of where you need a united response from both the tech sector and governments working together.

So in part, we’re using AI as a cyber shield, if you will. But then the other issue, and the one you’re really pointing to, is in a world where people are so dependent on technology, when governments, when countries are run on technology, is there a risk that technology gets turned off?

And I think we have to continue to advance the kind of international dialogue. So we have agreements between countries that will take that off the table, if you will, that will not make that part of a trade debate. If people want to apply tariffs, well, you’re increasing the prices, but the goods are still available.

And I think we need the kinds of steps, as we’ve been pursuing, say, with a country like India, for example, to provide assurance of supply. And I think this also really naturally involves more local control, local laws, local suppliers, local partnerships. All of these things have become part of a complex web that are just indispensable for how this technology is used in the future.

Ian Bremmer

Urgency of lack of international engagement on that last one right now, scale of one to 10?

Brad Smith

Glass is half full, it’s half empty, it’s a five, Ian, there’s so much more that could be done.

Ian Bremmer

Another question.

Audience

Namaskar. Hi. I’m Ishan Pratap Singh from the…

Global shapers community the New Delhi hub. I’m a 22 year old entrepreneur. My question is for mr.

Smith as well The and what would be the correct analogy for the example you used about? Railroads and power plants in context of AI and do you think the solution to fix that is to? Put universities and centers at the forefront rather than companies doing the job.

Brad Smith

The thing that I would highlight is that we’re what I would suggest in the seventh wave of technology driving infrastructure gone from canals to railroads to electricity to the telephone to highways for cars and airports for planes now we have AI it requires infrastructure, I think the Unlike say airports and highways But like the first four the private sector is investing globally.

That is good news Because that is helping to spread infrastructure around the world, but there are key shortages. We need to address Probably the most important in most countries is for people, but Kristalina was saying India actually is providing much of the IQ that is creating IP for the world But if you look at Africa for every one data scientist, there’s 14 data scientists in Europe That is a shortage that governments and universities can help me I would also say then we need more capital to be deployed where the market is not building these data centers Development banks development assistance can play a role but governments can help stimulate demand when the public sector adopts AI When countries consolidate demand by putting together say regional agreements to Make enable a data center to serve six countries and not one that will stimulate more investment We need to get the investment moving And I think governments are best served when they survey the market, look for the market failures and fill in the gaps.

Ian Bremmer

Last quick question, young woman in the front.

Audience

Hello, I’m Maria from Portugal, also Global Shaper. I’m an AI developer working on drug discovery. So you’ve mentioned the problem on entry level jobs, but you also mentioned education and has a follow up question.

Yes, education is very important, but the people are leaving universities and they need the entry level jobs. I don’t know if any of you have some thoughts on this question of, yes, we get education, but then how can the youth actually get these entry level jobs and what solutions can we actually bring for this?

Ian Bremmer

Kristalina.

Kristalina Georgieva

One important thing we see is to recognize that how people are prepared has to leave space to be ready for new skills and be able to fill applications where these new skills are a requirement. In other words, to learn and adapt rather than learn specific technical skills and stop there. It is not an easy problem and this is where we are telling governments that they have to invest much more thinking into helping communities, individuals, businesses to be prepared for the world of AI because it is here.

It’s no more a world of the future. And I want to say one thing about Africa. I have a very deep conviction that a way we can help Africa is if we finally fulfill one commitment that was made.

That is not very expensive. $15 billion to get every African citizen, business, and state institution connected to the internet. It is a step that would help tremendously.

Of course, there are issues of electricity. to be resolved, but it is not unresolvable if there is will and there is this strength of helping countries not to fall so dramatically behind. In my institution, we focus a lot on Africa.

We do one thing, which is digitalize all government services. It serves as an impetus to get countries moving. And actually, we strongly advocate for this issue of skills, be very proactive.

Don’t allow this to come so ahead than young people are on the street and desperate.

Ian Bremmer

I’m glad we ended with that. It’s a good note, and please join me in thanking our panel today. Thank you.

Thank you. Thank you all.

Kristalina Georgieva

Speech speed

122 words per minute

Speech length

1224 words

Speech time

599 seconds

AI is injecting resilience and strength into the world economy, with countries falling into three categories: those who make it happen, those who watch it happen, and those wondering what’s happening

Explanation

Georgieva argues that AI is a significant factor strengthening the global economy alongside private sector advancement. She categorizes countries based on their AI readiness using an IMF Index of Preparedness that evaluates four key areas to determine how well countries are positioned for AI adoption.

Evidence

IMF Index of Preparedness evaluates physical infrastructure, labor skills capabilities, diffusion, and regulation/ethics. Top three countries are US, Denmark, and Singapore. China has potent capabilities but ranks lower due to size. Saudi Arabia and India are in higher spectrum among emerging markets.

Major discussion point

AI’s Impact on Global Economic Power Distribution

Topics

Economic | Development | Infrastructure

Agreed with

– Ashwini Vaishnaw
– Khalid Al-Falih

Agreed on

AI is a transformative technology requiring comprehensive national strategies across multiple layers

Disagreed with

– Ashwini Vaishnaw

Disagreed on

India’s classification in AI readiness tiers

AI touches 40% of jobs globally on average, with 60% in advanced economies and 25% in low-income countries, creating an ‘accordion of opportunities’

Explanation

Georgieva describes AI’s impact on labor markets as dramatic and uneven, affecting different types of economies at vastly different rates. She warns this creates a widening gap in opportunities between advanced and developing economies, comparing it to an accordion that expands unevenly.

Evidence

Brad Smith

Speech speed

164 words per minute

Speech length

1625 words

Speech time

592 seconds

The US and China are competing on innovation leadership and global technology diffusion, while transatlantic ties remain crucial for AI development

Explanation

Smith identifies two key levels of US-China competition in AI: leading-edge innovation in chips and models, and the global spread of their respective technologies. He emphasizes that despite geopolitical tensions, the economic ties between the US and NATO allies remain fundamental to AI development and should be considered in policy discussions.

Evidence

Disagreed with

– Ian Bremmer

Disagreed on

Primary political challenges facing AI adoption

Generative AI is used by 25% of the global north population but only 14% in the global south, with the gap widening

Explanation

Smith highlights a significant and growing digital divide in AI adoption between developed and developing regions. He draws historical parallels to colonial-era infrastructure development that served extraction rather than local population needs, warning that AI could exacerbate global inequality without deliberate intervention.

Evidence

Microsoft diffusion report data showing usage rates. Historical analogy of colonial powers building railroads for troop movement and resource extraction but not building power plants for local populations. Specific mention of Africa having 1 data scientist for every 14 in Europe.

Major discussion point

Infrastructure Development and Digital Divide

Topics

Development | Digital access | Infrastructure

Technology weaponization is a critical threat requiring international agreements and local partnerships to ensure supply assurance

Explanation

Smith acknowledges that every technology can be both a tool and weapon, emphasizing two key concerns: AI-enhanced cyber threats and the risk of technology being ‘turned off’ as a geopolitical weapon. He advocates for international agreements and local partnerships to mitigate these risks.

Evidence

References AI being used to increase cyber threats and the need for AI as a ‘cyber shield.’ Mentions agreements with countries like India to provide assurance of supply and the importance of local control, laws, suppliers, and partnerships.

Major discussion point

AI Governance and Regulation Challenges

Topics

Cybersecurity | Legal and regulatory | Human rights

Agreed with

– Ashwini Vaishnaw
– Khalid Al-Falih

Agreed on

AI governance requires technical solutions combined with regulatory frameworks

Investment in skilling is more feasible for governments than data center investment and can stimulate private sector demand

Explanation

Smith argues that government investment in skills development is more cost-effective than infrastructure investment and can create market demand that encourages private sector data center investment. He emphasizes that skilling government employees and data scientists through universities can kickstart local AI adoption.

Evidence

Suggests skilling government employees as proven effective strategy, training data scientists through universities to build local applications, and notes that stimulating demand through skills development encourages private investment in data centers.

Major discussion point

AI Governance and Regulation Challenges

Topics

Development | Economic | Sociocultural

Agreed with

– Kristalina Georgieva
– Ashwini Vaishnaw

Agreed on

Skills development and education are critical for AI adoption and addressing labor market disruption

Ashwini Vaishnaw

Speech speed

150 words per minute

Speech length

899 words

Speech time

357 seconds

India operates across all five layers of AI architecture and should be classified in the first tier of AI-ready countries, not the second

Explanation

Vaishnaw challenges the classification of India as a second-tier AI country, arguing that India works comprehensively across all AI architecture layers from applications to energy. He contends that India’s strength in AI services and model development places it firmly in the top tier of AI-ready nations.

Evidence

The focus should be on return on investment through deploying lowest-cost solutions rather than creating the largest models

Explanation

Vaishnaw emphasizes that the economics of AI should prioritize practical ROI over model size, arguing that the fifth industrial revolution’s success will be measured by cost-effective deployment rather than computational scale. He suggests that smaller, more efficient models provide better economic outcomes.

Evidence

95% of work can be accomplished with 20-50 billion parameter models. ROI comes from deploying lowest cost solutions for highest returns. India focuses on AI diffusion and practical applications rather than largest models.

Major discussion point

Economic Measurement and Productivity Impact

Topics

Economic | Development

Khalid Al-Falih

Speech speed

136 words per minute

Speech length

Infrastructure Development and Digital Divide

Topics

Infrastructure | Legal and regulatory | Sociocultural

Agreed with

– Kristalina Georgieva
– Ashwini Vaishnaw

Agreed on

AI is a transformative technology requiring comprehensive national strategies across multiple layers

Sovereign AI models are necessary to ensure unbiased data and consistency with local values, as models developed in one country may not be relevant globally

Explanation

Al-Falih argues that AI models trained in one cultural context may not be appropriate for other regions due to inherent biases in training data. He emphasizes the importance of developing AI systems that reflect local values and cultural contexts rather than imposing external value systems.

Evidence

Concern about biases in datasets that models learn from. Need for models consistent with Saudi/Middle Eastern values. Importance of protecting value sets and character while gaining AI benefits. Models developed in Europe, US, or China may not be relevant for other countries.

Major discussion point

AI Governance and Regulation Challenges

Topics

Human rights | Sociocultural | Legal and regulatory

Agreed with

– Brad Smith
– Ashwini Vaishnaw

Agreed on

AI governance requires technical solutions combined with regulatory frameworks

AI applications in Saudi Arabia have demonstrated concrete results, such as reducing diabetic eye disease diagnosis time by 80% and chemical usage in desalination by 30%

Explanation

Al-Falih provides specific examples of successful AI implementation in Saudi Arabia across healthcare and environmental applications. These examples demonstrate practical benefits and measurable improvements in efficiency and resource utilization from AI deployment.

Evidence

Diabetic eye disease diagnosis time cut by 80%, saving thousands of clinical physician hours. Aqua Power uses AI cameras on wind turbine blades to detect approaching birds and shut down turbines. Same company reduced chemicals in water desalination by 30% using AI.

Major discussion point

Economic Measurement and Productivity Impact

Topics

Economic | Development | Infrastructure

Ian Bremmer

Speech speed

165 words per minute

Speech length

1614 words

Speech time

584 seconds

AI represents one of the most revolutionary set of technologies, fundamentally changing global power dynamics and requiring careful analysis of geopolitical implications

Explanation

Bremmer frames AI as a transformative force that is reshaping international relations and economic power structures. He emphasizes that AI’s impact on geopolitics is happening rapidly and involves massive investment flows, making it a critical factor in understanding contemporary global dynamics.

Evidence

References AI as involving ‘an incredible amount of investment’ and being ‘very fast moving,’ noting it’s one of the reasons the IMF upgraded economic outlook

Major discussion point

AI’s Impact on Global Economic Power Distribution

Topics

Economic | Development | Infrastructure

The transatlantic relationship and concerns about data centers represent more immediate political challenges than job displacement in AI adoption

Explanation

Bremmer identifies current political tensions around AI infrastructure and international cooperation as more pressing near-term issues than employment concerns. He suggests that community resistance to data centers and geopolitical tensions are the primary obstacles facing AI deployment today.

Evidence

References community opposition to data centers over water and electricity concerns, and mentions transatlantic relationship tensions in the context of AI and technology cooperation

Major discussion point

Labor Market Transformation and Skills Gap

Topics

Infrastructure | Economic | Sociocultural

Disagreed with

– Brad Smith

Disagreed on

Primary political challenges facing AI adoption

AI governance is moving much faster than regulatory capabilities, creating urgent need for feasible international cooperation mechanisms

Explanation

Bremmer argues that the rapid pace of AI development is outstripping the ability of governments and international institutions to create effective governance frameworks. He emphasizes the need for practical, achievable solutions given current political leadership and constraints.

Evidence

Notes that ‘this space is moving a lot faster than governance capabilities’ and questions what is ‘feasible’ with ‘most of the leaders that we presently have now’

Major discussion point

AI Governance and Regulation Challenges

Topics

Legal and regulatory | Development

Audience

Speech speed

165 words per minute

Speech length

Labor Market Transformation and Skills Gap

Topics

Economic | Development | Sociocultural

Agreements

Legal and regulatory | Human rights | Cybersecurity

Development | Digital access | Infrastructure

Human rights | Sociocultural | Legal and regulatory

Overall assessment

Summary

The discussion revealed strong consensus on the transformative nature of AI, the critical importance of skills development, the need for comprehensive national strategies, and the requirement for technical solutions in governance. There was also agreement on addressing the growing digital divide between developed and developing regions.

Consensus level

High level of consensus on fundamental principles and challenges, with emerging economies asserting their position as first-tier AI players rather than followers. The agreement suggests a maturing global understanding of AI’s implications and the need for inclusive, technically-informed approaches to governance and development. This consensus could facilitate international cooperation on AI governance frameworks and development assistance programs.

Differences

Different viewpoints

India’s classification in AI readiness tiers

Speakers

– Kristalina Georgieva
– Ashwini Vaishnaw

Arguments

India operates across all five layers of AI architecture and should be classified in the first tier of AI-ready countries, not the second

Summary

Georgieva’s IMF classification initially placed India in a middle tier among emerging markets, while Vaishnaw strongly contested this, arguing India belongs in the top tier based on Stanford rankings and comprehensive AI capabilities across all architecture layers.

Topics

Economic | Development | Infrastructure

Economic viability of large AI models versus smaller parameter models

Speakers

– Ashwini Vaishnaw
– Brad Smith

Arguments

Creating large models doesn’t necessarily provide geopolitical power, as smaller parameter models can handle 95% of required work more cost-effectively

The US and China are competing on innovation leadership and global technology diffusion, while transatlantic ties remain crucial for AI development

Summary

Vaishnaw argues that large AI models may be economically unsustainable and that smaller models are more practical, while Smith emphasizes the importance of leading-edge innovation competition between major powers, implying value in advanced model development.

Topics

Economic | Infrastructure

Primary political challenges facing AI adoption

Speakers

– Brad Smith
– Ian Bremmer

Arguments

The biggest political challenge is community concerns about data centers, electricity costs, and water usage rather than immediate job displacement

The transatlantic relationship and concerns about data centers represent more immediate political challenges than job displacement in AI adoption

Summary

While both agree on data center concerns, Smith focuses on local community resistance as the primary issue, while Bremmer emphasizes broader geopolitical tensions in transatlantic relationships as equally significant challenges.

Topics

Infrastructure | Economic | Sociocultural

Unexpected differences

Development | Digital access | Infrastructure

Takeaways

Key takeaways

AI is fundamentally reshaping global economic power distribution, with countries falling into three categories: makers, watchers, and those unaware of developments

The global AI divide is widening between the Global North (25% adoption) and Global South (14% adoption), requiring urgent infrastructure investment

AI impacts 40% of jobs globally, creating an ‘accordion of opportunities’ with advanced economies seeing 60% job impact versus 25% in low-income countries

Smaller, cost-effective AI models (20-50 billion parameters) can handle 95% of required work, challenging the assumption that larger models provide geopolitical advantage

Current political challenges focus more on data center community impacts (electricity, water, local benefits) than immediate job displacement concerns

Successful AI deployment requires investment across all five layers: application, model, chip, infrastructure, and energy

AI governance must adopt a ‘techno-legal’ approach, combining technical solutions with legal frameworks rather than relying solely on legislation

Sovereign AI models are essential for countries to maintain their values and avoid bias from models developed elsewhere

Skills development and education are more feasible government investments than data centers and can stimulate private sector demand

AI’s economic impact ranges from 0.1-0.8% of global growth, with the higher end representing significant productivity gains above pre-pandemic levels

Resolutions and action items

India committed to training 10 million people in AI skills through systematic programs

Saudi Arabia will continue building comprehensive AI infrastructure including data centers, sovereign data policies, and Arabic language models

IMF will continue working with researchers to better capture and measure AI productivity enhancement globally

Tech companies need to provide community assurances about electricity rates, water usage, and local job opportunities for data center projects

Development of technical tools to mitigate AI bias and enable proper judicial review of AI decisions

Continued investment in connecting African citizens to internet infrastructure as a foundational step

Unresolved issues

How to prevent the widening digital divide between Global North and Global South from becoming permanent

Addressing the entry-level job crisis for young people entering the workforce in an AI-transformed economy

Establishing international agreements to prevent technology weaponization and ensure supply assurance

Determining optimal governance frameworks that balance innovation with ethical considerations across different cultural contexts

Resolving community resistance to data center development while meeting AI infrastructure needs

Measuring and attributing productivity gains specifically to AI versus other organizational improvements

Ensuring AI benefits reach communities where AI is not currently present

Addressing the skills mismatch where some countries have high demand but low supply, while others have high supply but low demand

Suggested compromises

Public-private partnerships for AI infrastructure, as demonstrated by India’s common compute facility model

Maintaining investment optionality across multiple countries and technologies rather than depending on single sources

Balancing sovereign AI development with international cooperation and connectivity

Using government investment in skilling to stimulate private sector demand for AI infrastructure

Regional agreements to consolidate demand and enable data centers to serve multiple countries

Tech companies providing community benefits (electricity investment, water replenishment, local training) in exchange for data center approval

This philosophical observation about technology’s dual nature provided a framework for understanding AI’s potential for both benefit and harm. It acknowledged the inherent tension in technological development without being alarmist.

Impact

This comment introduced the critical security dimension to the discussion and led to conversations about weaponization, cyber threats, and the need for international agreements to prevent technology from being used as a geopolitical weapon.

Overall assessment

Explanation

Smith suggested that development banks and assistance could help deploy capital where the market is not building data centers, particularly in regions with market failures

Disclaimer: This is not an official session record. DiploAI generates these resources from audiovisual recordings, and they are presented as-is, including potential errors. Due to logistical challenges, such as discrepancies in audio/video or transcripts, names may be misspelled. We strive for accuracy to the best of our ability.