SATYA NADELLA: As we moved from GPT 2.5 to 3, we all started to see these emerging possibilities. It started showing the extended effects. We don’t train it to just code, but it’s really good at coding. That’s when I became a believer. I thought, “Wow, this is actually happening.”
Is there a eureka moment that makes you go all out?
It was that ability to code that led us to create Copilot. But the first time I saw what is now called GPT-4, in the summer of 2022, was an amazing experience. There is a query that I always use as a reference. Machine translation has been with us for a long time and it has achieved a lot of great standards, but it lacks the flair for capturing deep meaning in poetry. Growing up in Hyderabad, India, I dreamed of being able to read Persian poetry—especially the work of Rumi, which was translated into Urdu and then into English. The GPT-4 did it, in just one shot. It is not just a machine translation, but one that preserves the sovereignty of poetry across two linguistic boundaries. And that’s pretty cool.
Microsoft has been investing in AI for decades — don’t you have your own big language model? Why do you need OpenAI?
We had our own efforts, including a model called Turing inside Bing and powered in Azure and what you got. But I feel OpenAI is pursuing the same thing as we are. So instead of trying to train five different background models, I want to have a background, make it the basis for the background effect. So we cooperate. They bet on us, we bet on them. They do the foundational models, and we do a lot of the work around them, including the stuff around responsible AI and safe for AI. At the end of the day, we are two independent companies that work deeply together to pursue the same goal, with discipline, rather than multiple teams just doing random things. We said, “Let’s go after this and build something that really captures the imagination of the world.”
Have you tried buying OpenAI?
I grew up at Microsoft, dealing with partners in many interesting ways. In the past, we built SQL Server by working closely with SAP. So this kind of stuff is no stranger to me. What’s different is that OpenAI has an interesting structure; It’s a non-profit organization.
That often seems like a killer, but somehow you and OpenAI have come up with a complicated workaround.
They created a for-profit entity and we said, “We’re fine with that.” We have good trade partnership. I feel like there’s a stable long-term deal here.
Obviously, it’s set up for OpenAI to monetize your trade, just like Microsoft, but there’s a limit to how much profit your partnership can accrue. When you approach it, it’s like Cinderella’s carriage turns into a pumpkin—OpenAI becomes a pure non-profit. What happens to the partnership then? Can OpenAI say, “We’re completely non-profit, and we don’t want to be part of a commercial operation?”
