Thomas, thank you so much for joining us. I know it's been a really busy day coming out of Google Cloud Next. What is the clearest signal that Google's AI strategy has evolved and that you guys are ahead, not just playing catch up?
It's a combination. First of all, we have hundreds of customers talking about all the success they're having using our AI technology across virtually every industry and from all different parts of the world. And then we have a super strong slate of announcements. We've introduced over 3,000 new capabilities this last year across AI infrastructure, across AI models, across the tools to use the models, and a new set of products around what people call AI agents. So across every part of this portfolio, we have a lot of new exciting stuff. It's that combination and the fact that so many customers adopting it and the fact that at every part, if you look at infrastructure, cost of training, cost of inference, latency performance, we're the world leader. Across the model portfolio, Gemini Pro 2.5 is the world's leading model.
Okay, Thomas, we talk a lot about where Google is placed across every layer of the stack, in that sense unique. But talk a little bit more about the adoption piece. What can we look at to know that Google's winning more customers than your rivals and that's going to continue? Also the idea that generative AI is going to lead to top-line growth, not just cost-cutting.
So generative AI is helping people in retail improve online search on their website and in their catalogs, help media companies build recommendation systems to help people find the right movie to watch or the right information to kind of entertain themselves, help advertising — lots of people are using our tools to create content for advertising. We've seen people improve the precision with which new products can be built using our AI tools. All of these are for top line growth, not about cost optimization.
And where we're seeing adoption is the number of companies that are starting projects with us in AI. The number of companies from different industries that say, 'Okay, I really pick Google because you can really help me speed up how quickly I can take data from different systems, bring it into the cloud, and bring it into your platform, and then how well you integrate into my existing footprint.' Lastly, the ecosystem. We have all the major independent software vendors — Salesforce, ServiceNow, Workday, etc. — building integrations into our AI platform. We have an announcement tomorrow with something called our open API development kit that has support from 60 major software vendors to integrate into their environment. That gives the ability for people to get speed and value from using our AI tools. All those are unique. All those, and customers tell us the adoption they're seeing — we're the fastest growing.
Right. And so there haven't really been that many serious questions about demand. I think that continues to increase, but there have been questions around supply and the ability to build all the compute power. Now it's a little more uncertain amid the latest tariffs that the president and his administration have announced. Is Google accelerating infrastructure build out in response to potential cost pressures from those tariffs?
Well, on the tariff side it's so fluid I don't want to comment on it because it's changing by the minute. We're working similar to every other company to look at what we should do. But we've made a lot of enhancements to our models to make the cost of inferencing and the use of it scale. And the proof of it, we've seen over 2 million developers building daily, over 4 million developers building on our platform. We have over a billion API calls a week to our models. It's grown a hundred times in the last year. And all of it's because we've optimized the cost and the quality of the models. And as they get better and better, people feel comfortable using it much more widely.
Right. And there is sort of this worry about sourcing components and especially semiconductors, but of course Google uses its own custom TPUs. Does that help give you confidence that you're going to be able to meet needs if tariffs raise costs for the industry?
We co-engineer our models and the chips because there are a variety of different types of models. There's large models that need a different kind of infrastructure, that are what's called a
Mixture of experts model, there's a sparse model, and we have different chip form factors for these different kinds of models. Because we co-engineer the model and the infrastructure, we're able to do an exceptional job with quality and cost and latency. And the customers who use it love it.
You talked about sort of you guys are sort of dealing with the impact of tariffs in real time. And I just wonder, given all the market volatility we've seen over the last few days, within the executive ranks at Google, how have you guys experienced that volatility?
We've had a very strong team in moments like this. I think it comes down to how experienced is the team, how many cycles have you been through like this. And we've lived through many, many cycles like this. And we have a super experienced global team. Remember, we manage global infrastructure from the very time that Google was born as a company. And so, 25 years now we've been dealing with that. And so, we have a world-class team that both knows how to source, manage all our supply chain, but also operate it super efficiently. And so, we have, you know, with people working through it every day.
