Special Edition: Key findings from the Microsoft AI Diffusion Report

0:00There are problems that you cannot solve without using AI. And that's the part that we see so much opportunity for the world. I mean, it goes back to the history of every general purpose technology. The countries that grow the most over the course of years and decades are not necessarily the ones that produce it. It's the ones that figure out how to use it across the economy. That's what this report's about. What this dataset provides is a broad view of what is happening across platforms. We have a full view of what happens in AI today. That's Juan LaVista-Ferez, Chief Data Scientist and Director at Microsoft's AI for Good Lab.

0:38Welcome to a special edition of Tools and Weapons, as Juan and I review the results from Microsoft's newest AI Diffusion Report. These show the state of global AI usage as we look ahead in 2026, and there's a lot that stands out. First, while AI adoption continues to grow, there's a widening gap between the global north and global south. Second, some countries are sprinting ahead, like the UAE, Singapore, and the biggest mover, South Korea. And China's deep seek is gaining ground with implications for geopolitical AI competition. Our discussion about the new AI diffusion report, coming up next on Tools and Weapons.

1:23Juan, Happy New Year. Happy New Year, brother. It's great to be with you. This is our eighth year working together. And the reason we're getting together today is the new AI diffusion report. Yeah, we just published the report. This is the H2 report. We compared what happened on AI globally versus what happened in the first half of 2025. What this dataset provides is a broad view of what is happening across platforms. Microsoft has billions of devices out there. And we have an understanding of what is happening in these devices. And we have a full view of what happens in AI today. Everybody talks about AI diffusion, but I'm sure there's people who go, what are you guys talking about?

2:05People, when they think about general purpose technology, they talk about the diffusion of technology. In this particular way, it's how many people we have out there using AI. Like, it's about the use and adoption. Because AI is a general purpose technology like electricity, meaning ultimately it will touch over time and will improve every part of the economy. We live in a country where almost 100 % of the people, both at work and home, have electricity. Yes. We wouldn't be comfortable looking around and saying, oh my gosh, 80 % do and 20 % don't. We're far away from that when it comes to AI. There are problems that you cannot solve without using AI.

2:51And that's the part that we see so much opportunity for the world. Well, and one of my favorite stories was the one you told, literally having AI in the rainforest, constantly fueling devices to be able to measure biodiversity, not just with cameras that can see, but with microphones that can hear and AI that can process. There's 200 ,000 conservationists out there, and what we saw is a huge effort for them to collect data. And without data, they cannot do their job. They don't understand the species. For them, it's a game changer. One of my favorite projects is about a leading cause of blindness.

3:32The world does not have enough ophthalmologists, and in order for them to do screening, we need to have technology to help them. AI is there, not just a solution, but unfortunately the only solution we have. History tells us that the countries that grow the most over the course of years and decades are not necessarily the ones that produce it. It's the ones that figure out how to use it across the economy. That's what this report's about.

4:03Let's dive in. Let's talk about some of the stories that the numbers are showing us. AI in general, we are seeing it has been the fastest technology in adoption in the world, like even to social networks. We now have over 1.2 billion people around the world using this technology. And the global north grew 1.8 percent. The global south grew only 1 percent, which means that there was already a big gap between the two. Now that gap is widening. The global north is almost at 25 percent. It's at 24.7. Almost a quarter of the population in the global north is now using generative AI. In the global south, it's under 15%.

4:43It's 14.1. You look at those numbers and you say, wow, at this rate, another year or so, we may see the global north at twice what it is in the global south. Yes. That's something that will likely happen. And that's sobering because, in a sense, the economic divide between north and south on a planetary basis, I think, was in some ways created because of the divide in electricity use. And then the question is, well, will AI be something that can help people close this gap or will it widen that gap? The future is yet to be written, but right now we're going in the wrong direction. Yeah. And what is interesting to me is that if you look at the global north, majority of the country has access to internet.

5:33We don't see that in a lot of countries in the global south. So which means that a lot of these countries will even get to a point where there's no more growth for them. Yeah, we've created a leaderboard, looked at where countries rank. First half of 2025, the global leader was the UAE, the United Arab Emirates. It was number one. Singapore was number two. They were the two countries above 50 % usage. Now, the UAE has widened its lead over the course of the year. Globally, use went up 1.2%. UAE, it went up by what? 4.6%. per set. And this is something that is also kind of not expected. Like those countries that are the ones that have more share, it's more difficult for them because majority of the share, like usually these follows a curve.

6:21We're not seeing that in UAE. It's like other than South Korea, UAE is a country in the world that has grown more share. You were in the UAE in November. I was there in November. Why is the UAE doing so well? What I saw there was like kind of everybody's talking about AI. And everybody was talking about AI in a very positive way in many ways. That's something that I haven't seen in other countries. You have the government behind the fact that people need to see AI as an excellent tool to improve productivity, for example. When we went to the Department of Health at Abu Dhabi and then their command center, you could just see how AI was being used to make sure that ambulances got to a patient faster, could be redirected through traffic faster, to take a patient to the right hospital, and if a patient was likely to need to be hospitalized, to a hospital where there was going to be a bed available.

7:17That's called saving lives. But I also saw the trust, and I think you did too. Yeah, that's the part that I saw is like people would trust like, hey, this is something that it makes sense. It will help me do my job. It will help me increase the productivity. It was a very positive conversation.

7:39There was also one big mover. If you look at the leaderboard, the big surprise really was South Korea, wasn't it? It moved from, what, 25th in the world to 18th in the world? That's a big jump. How did it do that? Yeah, that's the biggest jump that we saw. There are three things that I think are different. The first one is similar to UAE. You have a government that is investing a lot in AI. It's talking a lot about AI. And I think that plays on the trust side. The second aspect is that in the early versions of JGBT, JGBT 3.5, it wasn't great in Korean. When OpenAI released 4.0, two things happened.

8:19First one is they bring the capability to share images. That goes viral. And it goes viral across the world, not just in Korea. But in Korea, we see a bigger adoption. And 4.0 was also much better in Korea. So I think that helped people understand that you could use it to generate Studio Ghibli, but you could also use it to answer questions. And I think that that virality helped. The difference between South Korea and the rest of the countries is in South Korea, that remains stable. Typically, when something goes viral, it goes up and then goes down. In the case of South Korea, it goes up and the amount of people using it remains.

8:56It stayed. Yeah. In some ways, it's the story, again, governments investing, it probably builds trust. Things are going viral, meaning people find it useful. They enjoy using it. But then I think this third dimension, to really be adopted by a broad population, it has to be good in the language that people speak. GBT 3.5, yeah, it wasn't great in Korea. By the time it got to 5, it was really quite, quite good. Yeah, there is product truth. And definitely that's something that we see today in South Korea. I suspect for Americans who are watching this or listening to us, they're like, wait, what's going on?

9:36I thought the U.S. was the global leader. That's what we hear all the time. We're number one. And when it comes to AI infrastructure, when it comes to frontier model innovation, the U.S. is. It is the leader in what you might call leading-edge innovation. But one of the reasons we're so passionate about diffusion is because as we study history, what we see is that if it's a general purpose technology, you don't have to be the leader in producing it to be the leader as a country in putting it to use. And if you can be the leader in putting it to use, you'll probably grow faster than other countries.

10:14But I said last month, well, but when it comes to AI adoption, AI diffusion, we're number 23. as a nation. That was from the data based on the first half of the year. The U.S. went down one. Now we can say we're number 24. Yeah. So the U.S. grew 2%. We see a lot of use of AI. We see a lot of subscribers. But yeah, on a per capita basis, it's not doing as well as many other countries in Europe and even in countries like UAE or Singapore. When I was in Ottawa last month with Prime Minister Carney. And he turned to me and said, hey, I've seen the data. Canada's number 14. How does Canada move up the ranks?

10:58What should it do? And I said, well, look, the history is a really good guide here. You should do one thing above all else, invest in skilling. That's another part of the UAE story. The UAE was not only the first country in the world to have an AI minister, but the very first thing the AI minister did in 2017 was say, I'm going to skill the government employees so that they know how to use AI. The government starts to use AI. That builds adoption. It builds trust. It all fits together. Now, you put that together, and you see the U.S. hasn't yet reached 30 % when the UAE is over 60. And we're not yet, I don't think, seeing investments at the national level, either collectively, you know, company by company, state by state.

11:48It's not yet accelerating in the U.S. or Canada or Western Europe the way it is, say, in the UAE or Singapore. Even in the case of South Korea, they actually have invested a lot in making sure that you have curriculum, AI curriculum as part of high school, for example. Also, you're a resident, K-12, on making sure that people have the skills. I think that the investment that you need for society to skill the AI population will be much smaller than doing, for example, for COVID. That's how I think there's an opportunity there. Right now, you see governments willing to invest large amounts of public money, taxpayer money, especially in, say, the European Union, to build out infrastructure.

12:31When I was in Portugal in November, they were putting together a bid to get$10 billion of European Union investment in AI infrastructure. And when we were there, we announced a$10 billion Microsoft investment in AI infrastructure. What people are not figuring out is what would be a smart way to spend$10 billion to skill people. And you look at, say, a country like France, widened its lead over the US, and the private sector is investing in infrastructure. If we want AI to make its way into education in a smart way that really equips people with the skills they need, that's where we're going to need more public sector support.

13:15And when you look at this list, majority of these countries in the top, at least in the top 20, all of them are very similar. So something is different in these countries. That's something that I would love for us to spend time is like going deep. It's like, what is the difference between countries? And usually likely skilling will play a role there. I think it's got to be one of the most important AI topics for 2026.

13:41The big story of 2025, when we look back at the past year, in some parts of the world, it was about DeepSeq. It was about Chinese open source models. What's the pattern that you see in the data there? When DeepSeek started in January, first, it was kind of the only new model that was now playing at the frontier. But it was also different in the sense that it was open source. Anybody could potentially use it. And their service was free, meaning that not only the model was open-weights, not necessarily open source, but the application didn't require a credit card. But the rest of the models, whether it's Gemini or OpenAI, JGBT or Anthropic, they usually have a free version.

14:27But this was different because it was at the same rate, but free. When we look at the data, you have big disparity between countries like China, Russia, Belarus, or Cuba. All of these countries have very high use of DeepSeek. And in the global south, we see something different. On one side, Latin America has very low usage, but Africa, actually, you have a lot of countries that have higher than 20 points of share. A product that didn't exist last year. So definitely the PSEC has been the revelation of this year from a new model perspective, and it's changing the dynamics in the world, I would say.

15:07From a U.S. perspective, or say people who are U.S. policymakers, one of the lessons from this report is the real competition is also about diffusion. U.S. should feel good about the global north. But in Africa, that's the place where Chinese AI is accelerating. American AI is not. Matt, if people in the U.S. Congress want to think about the future, it's probably worth spending a little bit of time thinking about the continent that's growing the most. Yes. Where the population is the youngest, that is going to have such an impact over the next quarter century. Yes, certainly. And when you think about the next 50, 60 years, we have to remember today that in countries like Nigeria, you have more babies born in a year than you have in Europe.

15:55Correct? So that population dynamic is changing. this continent will become even more relevant. And that's where we see, again, outside China, Russia, Belarus, or Cuba, that's where we see the biggest market share for DeepSea. As we look to 2026, the weeks and months ahead, any early predictions of what you think we might see? We will continue to see growth of AI adoption. I don't think this is going to stop. Unfortunately, I think it's going to be easy to predict that that gap between the global north and the global south will continue to grow, too. But, yeah, I think that what we'll see will be better models.

16:39Hopefully, the growth in the global south, people will have adoption, too. But, yeah. I think countries, understandably, naturally, maybe even reasonably, are thinking about AI as a race, as competition. The first question that I hope more people will think about, especially if they're in positions of, say, government leadership and responsibility, is what the race is all about. It's not just about innovation and infrastructure. Don't think less about those, but think more about diffusion and adoption. Second, know your numbers. Know where you stand in the world. You may be in a different position than you think, and you'll never know unless you take the time to understand your numbers.

17:24Third, I think every government leader should ask the question that Prime Minister Carney asked me. How can we do better? How can we rise up the ranks? And then finally, if there's one thing we've learned in 51 years almost at Microsoft, don't just learn from yourself. Learn from your competitors. Always learn from the leaders. If the UAE is leading because the government's adopting AI, the public sector is helping create great uses for AI, people are investing in skilling in AI, and they're doing what it takes to build trust in AI, that seems like a pretty good combination. Do you think there'll come a point in the future, decades, years ahead, where people are going to say, no, we need to have 100 % AI adoption, AI diffusion in order to feel good about opportunities for everybody?

18:19That's a great question. I think that when we look at the trend, I expect the numbers will continue going up. When we think about how people can leverage this technology from healthcare to industries to their own productivity. is like, for example, when I look at my team, like I want everybody to be using this technology. Like I think it's great for me. I would love for people to understand that this is something that gets people more productive. We'll have more to talk about. I know you're going to be measuring even more. And for those of you who joined today, thank you. You've been listening to Tools and Weapons with me, Brad Smith.

18:56If you enjoyed today's show, please follow us wherever you like to listen. This episode is also available in video on the Microsoft YouTube channel. Our executive producers are Carol Ann Brown and Brent Christofferson. This episode of Tools and Weapons was produced by Mike Morrisrow and Lexi Swanson. This podcast is edited and mixed by Trifilm, with production support by Sam Kirkpatrick at RUN Studios. Original music by Angular Wave Research. Tools and Weapons is a production of Microsoft.