La race for artificial intelligence is entering a new unexplored and expensive phase. The 2025 was the year of mega-deals: billions invested in data center, cross-fertilization of key companies and a race for to grab the best and brightest minds. One of these minds is Mustafa Suleyman, which has been in the last 18 months head of artificial intelligence at Microsoft. Suleyman made his name by co-founding DeepMind, acquired by Google in 2014, which later produced the artificial intelligence system that defeated a world champion at Go.
In ecosystem, its ability to break new ground in the industry was until recently limited by the terms of a agreement with OpenAI, but the recent revision of that agreement now allows Suleyman to make his new goals public. He talks about it in this interview with BloombergThe new frontier of artificial intelligence is “superintelligence,” which Suleyman prefers to call “humanist superintelligence” to ensure the survival of our species and the prosperity and well-being of all. And he talks about “red lines” Not to be crossed. The first uses of superintelligence will be in the medical field and in renewable energy. It will give us the ability to work less, if we want, and will produce abundance.
What uses of artificial intelligence in your life might not have been exploited by the rest of us yet?
Yesterday I stayed up late watching a movie and afterward I added something to a table I created in Co-pilot, which basically records all the movies I love, sorted by date. I add my personal notes and get a link to the movie poster. I could keep asking, "What similar movie?" You can ask your AI to perform virtually any cognitive task, just as you might ask an assistant to organize your life. The more complex, creative, and challenging the task you give your AI, the better.
Have you ever used AI to perform autonomous tasks? Did it book tickets for you or buy you a gift?
We're still experimenting. It can do this. It doesn't always work, but when it does, it's the most magical thing you've ever seen. Basically, it types things into your browser, clicks buttons, opens new tabs. It can look at your history and personalize your purchase or response.
What mistakes have you made that have caused you problems? Did you buy a gift for the wrong person?
Well, it might buy the wrong thing, but you can intervene. And it will always ask your permission before taking the next action, so it's pretty safe. It's a strange thing, technology. It's magical and amazing, and it's always a small step forward. I know it will work in the next six or twelve months, or at worst, eighteen months. It's already superhuman. It's very possible that next year we'll be able to buy Christmas presents using an autonomous AI agent.
Sam Altman wrote that OpenAI would go beyond artificial general intelligence—AGI, or artificial intelligence that rivals human capabilities—to create superintelligence. The term is now dominating the conversation in Silicon Valley. In June, Mark Zuckerberg restructured Meta's AI division, transforming it into Meta Superintelligence. You introduced Microsoft AI's Superintelligence Team last month. What does "superintelligence" mean to you?
Superintelligence represents an artificial intelligence system capable of learning any new task and achieving better results than all humans combined, at any task. This is a very ambitious goal and, at the moment, involves a high risk. It is very uncertain how we could contain and align such a powerful system. My preferred definition is that of a humanistic superintelligence, always by our side, on our team, aligned with human interests. Until we prove it will remain secure, we will not continue to develop a system that could escape us. Yet, I believe it is an innovative position in the industry right now. This is our position. Microsoft is a company that has been around for 50 years. It is very careful. It is highly reliable: 90% of S&P 500 companies rely on us for email, operating systems, and daily productivity. We have this reputation because the company has been careful. We will continue to be careful, and define a vision of humanist superintelligence is part of this program.
What does this mean for your competitors, some of whom you work closely with, like OpenAI? Does it mean they're the Wild West and you're the moral one?
Everyone must decide what they stand for and how they operate. I don't want to judge how they are operating right now. I don't see any evidence of large-scale mass harm. I've clearly identified the factors that increase the level of risk. We must address them cautiously, with greater transparency and oversight, with greater government involvement, and proactively declare how close we are to these capabilities. Containment and alignment are necessary prerequisites, red lines. I believe everyone in the industry must subscribe to this idea. No one wants to cause mass harm. Even if we don't all agree, we are all committed to the survival of our species and, I hope, to the prosperity and well-being of all. It's a great responsibility. This is a time when the decisions we make can have very long-lasting consequences.
Will the first uses of superintelligence be in the medical field?
I think so. This is probably the most exciting application of superintelligence. We now have systems capable of diagnosing any rare condition in the literature, significantly better than human performance, at lower costs, with fewer tests, and with greater accuracy. We're currently submitting everything to independent peer review, and soon there will be clinical trials. I also want to do more in energy efficiency and battery storage, developing new compounds for renewable energy. I believe artificial intelligence will truly transform the energy sector. Everything is really, really, really exciting. Technology should improve our lives, make us more comfortable. One day, I believe it will help us live longer. It will give us the ability to work less, if we want. It will produce abundance.
So will humans have to work less?
It's inevitable that, at some point in the next 20 or 30 years, machines will be more capable than humans at performing most jobs; this could happen much sooner. We must decide as a society what our purpose is. We must be very careful about the pace of introducing new machines, because we must ensure that replacement is offset by a mechanism for financing and supporting people during a radical transition. A universal basic income is inevitable and highly desirable. We already live in a world of abundance, it's just unevenly distributed. Value isn't just manifested in atoms—food, cars, physical objects. It manifests in digital assets—ideas, knowledge, intelligence.
Do you talk to your colleagues, like Sam Altman, about the risks of artificial intelligence?
Yes. Everyone in the industry does it. There's definitely a group of CEOs where we all know each other very well, like Sam, Dario [Amodei] from Anthropic, and Demis [Hassabis]. Overall, everyone is genuinely committed to trying to find the right path. It's also very competitive. Sam altman He's brave. He's expanding his data center fleet very aggressively, faster than anyone else in the industry, and if he succeeds, it will be a resounding success. He could very well turn out to be one of the greatest entrepreneurs of our generation. He's certainly achieved great things.
OpenAI is investing huge amounts of money. Is it a gamble? Will it pay off?
ChatGPT is one of the best products we've seen in a generation, and that speaks for itself. At the same time, they've signed commitments of over $1,5 trillion to build data centers over the next five to ten years, and their revenues are still far from that goal. So they still have a long way to go, but they're a very talented team. I'm absolutely confident they'll make it. All companies in the industry are increasing capital expenditures to fund AI.
How would you describe Elon Musk?
I imagine a bulldozer. He has superhuman abilities to bend reality to his will and a truly incredible track record. Somehow, he almost always manages to accomplish the seemingly impossible. He probably has a different set of values.
Do you think more regulation is needed?
It's a controversial thing to say in Silicon Valley, but I think regulation is necessary and has improved most technologies. People forget this, but cars only work because we have driving lessons, emissions regulations, streetlights, and speed limits. That's what regulation is for when it works well. We need more regulation, but I don't think we're at a stage where we need emergency regulation. Some of the regulations proposed in Europe are currently being negotiated to be scaled back, with the EU AI Act. I think that's a positive thing. It shouldn't be criticized. It's the process that works: the regulator gathers feedback and verifies how things work in practice. This should be appreciated.
You've renegotiated an agreement with OpenAI. You now have the freedom to develop AI independently. How will the relationship with OpenAI work in the future? Until now, you were collaborators; now you'll also be competitors.
It's a complicated system, but the bottom line is that until a few weeks ago, Microsoft was contractually prohibited from independently pursuing artificial general intelligence (AGI) or superintelligence. Microsoft has a license to everything OpenAI develops, until 2032. But OpenAI has decided to acquire more computing resources and buy them from other vendors: it now has agreements with SoftBank and many others to build more data centers than Microsoft wanted to build for them. In exchange, we have the right to develop our own AI. We are now hiring a superintelligence team and advancing the development of our own AI.
Does this mean you are ready to match the sums Meta has spent so far, such as packages of 100 to 200 million dollars?
I don't think anyone is imitating them. Zuck has taken a particular approach that involves hiring many people, rather than building a team. I really don't think that's the right approach. What I've done at Microsoft over the last year and a half is gradually add people who fit the culture, have the skills, and work well with the rest of the group, eliminating those who don't, and paying close attention to detail. We're building a team, rather than a collection of individuals.
Does Mark Zuckerberg, or Zuck, as he calls him, make your life more difficult?
No, not at all. He's building a 2-gigawatt data center, which will likely cost him a couple hundred billion dollars over the next two or three years. He publicly told Trump that he will spend more than $600 billion on data centers over the next three years. Fortunately for Microsoft, we have 33 gigawatts of computing capacity in our fleet. Our primary business is building data centers and providing them to third parties. We can use them for our own internal training or for inference, selling them to third parties. We're very well-positioned for that. We're protected from that risk.
What role does AI play in your personal day?
I'm really proud of many of Copilot's use cases. Many people use it for companionship, therapy, and to help them make difficult life decisions. It's given me high-quality access to information and emotional support, and it helps me stay organized. At the end of the day, when I'm in the car driving home, I talk to Copilot for 10 minutes about something that's been bothering me or frustrating me. Maybe emotional support is a bit strong, but it's like chatting with a friend: you talk about what went well and what didn't. Copilot now remembers most of what I've said and will personalize its responses based on my needs, referencing, for example, something I said last week, or a trend or pattern. It's really helpful. I feel refreshed after a conversation. It's like a weight has been lifted off my shoulders. It's like a friend, a therapist, a family member, almost.
People go home and perhaps no longer need to talk to others, because they have said everything and received back what they needed. Or not?
Or the flip side is that they don't have to vent on their partner or best friends. I still call my best friends every weekend and have a chat. In fact, I've even deepened some of my relationships with my friends.
Do you think AI could have had this conversation with her and do a better job? How can AI help journalists?
AI could have prompted other questions, too. I run MSN at Microsoft; it's one of the largest news sites on the planet. One of the things that excites me most is how AI-powered journalists can revitalize local news. Imagine hundreds of thousands of AI-powered journalists able to call people on the spot, verify eyewitness footage, conduct interviews, and stitch it together into small edits—and not just for big national stories, where the investment is justified, but at the local level, to provide accurate and factually reliable information.
