Pinterest CEO on AI Roadmap, China’s Emergency Nvidia Meeting, AWS’ New AI Model | Dec 10, 2025

Thanks for having me, Akos.

Yeah, I think so much of the discussion around AI has been about who's building the latest largest proprietary model. And I think there are a couple other really important trends that are happening that aren't being talked about enough. One is open source. As we published, we're seeing that we can take open source models and fine-tune those and get similar performance to the very best proprietary models at less than 10% of the cost. And I think that's really really important because open source has a democratizing effect that has been true for building software for decades now. And I think that's really important for how we look forward as to how many many companies can build with AI. If you can get similar capabilities for less than 10% of the cost, that's going to have a really democratizing effect, really spur a lot of innovation, creativity. And so we're pushing our work on that to share with others just how good that is because we think it's really important there's a thriving open-source community for AI. And then there's a second trend around compact fit-for-purpose models that I think also is about you don't need the very largest models in the world to do every single task. And that's also a way that you can create cost effectiveness in AI as well.

Two things. One, there are large-scale open-source models available now that you can take and fine-tune them for your purpose. And what we found for our purposes is that we could achieve 90% lower cost at similar performance on the things that we needed to do.

Well, yes. But I think this is a generalizable point because that's about the large generalized models. There's a second point which is related to this, that we are also building our own in-house models for specific tasks. And so these are two different tactics that can help companies get a lot more out of AI at much lower cost. And this is really important right now because nearly every CEO that I talked to, for example, I was last week at the New York Times DealBook. And so many of the CEOs that I talked to there said they've invested a lot of money in AI and buying these off-the-shelf proprietary software solutions, but they're not seeing the return on investment that they need. They're spending a lot, but they're not yet getting the savings that they had expected, or the spend is significantly offsetting the savings that they were getting. Companies are finding that they can leverage open-source to get much better efficiency. This has been true in software development for decades. In fact, many of the largest companies out there wouldn't exist if not for the fact that they have been able to build on open-source. It was much cheaper than proprietary databases or proprietary operating systems. Things like Linux were really really important to the building of a lot of these trillion dollar market cap companies. Well, the next batch of companies, they're going to need to have really great open-source AI available to them to have these cost efficiencies. Otherwise, the proprietary software may collect all the value without others being able to go build and create value from that as well. So we think this is really important, which is why we're sharing our findings.

Yeah, I mean the DeepSeek moment was a huge moment. And I think so much of the DeepSeek moment the conversation was about how they had done that on less than the latest GPU chips, the sort of theory of constraints of how they leveraged those constraints to find cheaper ways to build those models. That was really about the chips, but there were a lot of techniques they used that are now going into the open source community that others can use to build their own models. So we're using some of those techniques for our own models that we're building internally. But then you've had other really good open source models emerge, like Qwen for example, that is putting out really powerful models. But I think that DeepSeek moment, so much of the focus was on the chips and lower cost chips. But I think the bigger moment was that you had an open-source model that was rivaling the large proprietary models. And now you have that with Qwen. You see others coming out, Mistral, you're seeing an open-source community really start to develop that can compete. Open source is at the table and it is competing and doing so at really effective cost to performance levels. And that's important not just from a cost perspective. I think it's also important that so many people refer to these models as this model came from this country or that model came from that country. The really important thing about open source is that open source belongs to everyone. It's not controlled by anyone company. True open source is not controlled by anyone company. It's not controlled by any one nation state. And if you want open AI to be in the hands of the many rather than the few, then open source is really important to that. And again, that has been really important to software development for decades now. And I think it's really important that that continues in the world of AI. And it doesn't need to be instead of proprietary models. Open source software has been huge for decades. It doesn't mean that proprietary software no longer exists. It just means that open source software is a really important component of a thriving ecosystem overall alongside of proprietary models and can compete well with those proprietary models and it's a good check and balance on those proprietary models.

Well, we're applying the gas. We've effectively turned Pinterest into an AI powered shopping assistant over the last few years. We've had nine straight quarters of record high users. More than half the platform is Gen Z. Shopping is the primary reason they come to the platform. And if you ask Gen Z users why they come to Pinterest, they'll say things like, 'Well, Pinterest just gets me.' And that's us using our AI models tuned off our unique curation signal to give really great personalized recommendations and agentic style experiences that help guide users through shopping journey. So we're really doubling down on that. We've just launched our latest Pinterest assistant going even further with that. But we're building shopping, we're building AI experiences that don't replace shopping or automate away shopping. I sort of see that folks are doing that, trying to build shopping for people that hate shopping. We're building shopping assistance for people that love shopping. And they want an assistant to actually help them immerse in that journey. So that's where we're doubling down and hitting the gas. To your point, the place where I wouldn't say we're hitting the brakes, I'd say we're just making sure that we use AI responsibly. The first is tuning AI for positivity. This is one of the primary reasons I joined Pinterest from Google nearly three and a half years ago, is that I wanted to prove there was a more positive alternative to the business model of social media that so much of it had engagement via enragement at the core. A big part of that, how did social media become negative? AI got put in charge of what you see on social media more than a decade ago. It was just earlier forms of AI.

Well, avoiding slop or avoiding other negative things. When AI was asked to maximize your view time on social media more than a decade ago, the AI figured out you look longer at the things that trigger you, whatever your triggers are. And so we set out to tune AI for positivity. If AI could be used to keep you glued to a screen with engagement via enragement, why can't we ask the AI to make sure you leave the platform feeling better? And that it's a positive contributor to your emotional well-being. And we've been able to prove that out. So that's a place where we're making sure we don't have a race to the bottom on just trying to keep people glued to a screen doing things that are addictive rather than additive. And so we try to focus on additive. Another place is on trust and safety where we're using AI to combat bad content. You asked about AI slop. Well, there's been a huge increase in the ability to create content. The vast majority of that are people that are having fun creating good content. There's been a democratization of people's ability to express themselves. That's a good thing. But also mixed in that, like any technology, generative AI can be used for good or for bad. And so you also have bad actors that are trying to create things that are spammy or not helpful or harmful.

Oh, absolutely. And what I would say with AI generated content, this whole discussion of AI slop, a couple things I'd say. One is that it's not about is AI content good or bad. It's about some AI content is good and some AI content is bad. And how do you parse that? And how do you give the user control over what content they want to see?

That's exactly right. So two things that we're doing. The first is that we are labeling AI content. So the user knows when it's AI generated and we're using industry techniques to detect and label when something is AI generated. But not everybody's labeling that. We are further along in the industry, I think, in making sure that we label. And we've made a choice to label when we can detect that it's AI generated content. Secondly, we're giving the user choice in letting the user decide when they want to see less AI content. And so that's very different than other platforms. And I think part of why platforms may be avoiding this is that they're seeing that the AI generated content is really engaging. It keeps people looking for longer. But for us, we're not trying to keep people looking for longer. We're trying to help people do things that make a positive impact in their real life, even when that means something off their platform, like going and buying a new outfit or going and redesigning a room. And so we consistently give our users agency, both the option on the platform and then they have the option of when the AI generated content is helpful or not. And we're personalizing more and more to understand for each user what's helpful and not. Some of these things, to give you some examples of when is it helpful versus when is it not. There's some content that's just inherently bad, that would be created by bots or there's spam content that we use AI models to just get that stuff off the platform. But there's other content. We've always said beauty's in the eye of the beholder, or one person's trash can be another person's treasure. Something that one person thinks is art, go back to modern art movements. Traditionalists would have said, 'Oh, that art.' This is the thing, it's everyone's opinion really what they want to see. So I think that choice is interesting.

Well, advertisers have behaved very consistently. Advertisers will go where consumers are, and advertisers will really go where consumers are actually making purchasing choices. For our business, wherever purchasing choices are happening, advertisers are going to go there. And clearly AI assistants and chatbots are a place where that is happening and where advertisers are certainly going to go as they have the ability to do that. But it's not a zero-sum game because at the end of the day advertisers are looking to get incremental purchasing. And for us, even as you've seen the rise of chatbots and assistants, I think over the last three years ChatGPT has put on roughly 800 million users. Well, at the same time, we've had nine straight quarters of record high users. More than half of our platform is Gen Z. And I would guess that probably nearly every one of those Gen Z users on our platform, they know about chatbots and they view them, but they see us as something unique and different from that. And so there's room for both of these to exist. And that actually mirrors the way search has been playing out for decades. For decades you've had general purpose search where like Google is a winner of general purpose search, but then you also had vertical specific searches that coexisted with that. More product searches start on Amazon than start on Google. Travel searches start on Booking or Expedia rather than Google. And while you did have the general purpose capability, you could have a vertical specific capability where you usually could go deeper, whether it be on Amazon for shopping or Booking or Expedia for travel. And I think similarly in this AI world, you're going to have general purpose AI akin to what general purpose search was. You'll have a small number of large winners in that. But then also you're going to have fit-for-purpose tools that for a specific task, just as I was talking about the models, you can get fit-for-purpose models that perform that task better. I think you're going to have companies and products that do that better. And that's been my thesis from the beginning with Pinterest, that for visual search and shopping we can really focus on that and outperform. One last thing to share with you on that. On our last two earnings calls I've shared, our latest multimodal visual search models outperform the leading proprietary models by more than 30 full percentage points on the relevancy of their shopping recommendations. So not for anything you can ask and not for all of human knowledge, but specifically to shopping, outperforms on the relevancy of recommendations by 30 full percentage points because of this issue of the smaller compact fit-for-purpose models with our unique signal around user behavior and curating, we can outperform. And that's why those Gen Z users that are flocking to our platform will say things like, 'Well, Pinterest just gets me,' because we get the personalization right through those shopping recommendations.

Well, I think one of the things you're alluding to that we have done is that we've focused on making sure that the ads are great content for our users. When the user is in a shopping mode, which is the majority of users on Pinterest are there to shop. I talked about winning with Gen Z, more than two-thirds of Gen Z are coming to Pinterest to shop. An Adobe study came out that said 39% of Gen Z thinks of Pinterest as a first place to go search. 70% of them see Pinterest as more personalized. And so those are examples of where they see that we're providing a really good fit-for-purpose solution on shopping. But it also means that when somebody's shopping, as long as we show them the right product, it doesn't matter so much whether it's an ad or not an ad. Did you show them the right pair of shoes that they're actually looking for? That's what matters. And so for others, I think this will be a real question: when you have commercial intent, the ads can be great. When the user doesn't have commercial intent, the ads may not be so great. And I think with a lot of the chatbots, there's going to be some commercial behavior in there. There's going to be a lot of behavior in there that's non-commercial. But us, our platform is primarily about shopping. And so that makes it so that we can really deliver a great experience for the user. As we've delivered those nine straight quarters of record high users, we've also consistently deepened engagement per user, which means we're making the ads relevant content for them. They're helpful to the user. But that's also great for the advertiser because that means the advertiser gets to meet the user in a moment where the user actually wants to see their ad. In some other place where the user might be researching or doing other things, the user may not want to see their ad in that moment. And so here we can align the incentives of the user and the advertiser because of the shopping context.

Well, you're absolutely right. It's more than just implementing AI in the core platform. In fact, we've had a major reinvention of the platform over the last 3 years. So we are in the middle of that reinvention. Three and a half years ago, Pinterest was declining users and was focusing on short form video like everybody else and had lost differentiation and relevance. As we focused on turning Pinterest into an AI-driven shopping assistant, that has led to the resurgence of the platform. That has been a massive reinvention of the platform. And we did that not just by implementing AI. It was that there was a really unique behavior on Pinterest that we've really doubled down on in terms of human curation. I think so much of the discussion about AI has been about how does AI just automate away all the things that humans will do, and we'll all just go live on a universal basic income and have no work to do. It's a very interesting life, I think most people wouldn't think that's a very interesting life. We're focused on how do we make the AI additive for people and truly helpful to people so that people can be more productive, get a higher quality of life. That human curation on our platform is a good example. When you get that 70% plus of Gen Z that sees Pinterest as more personalized than other places to go search, part of that is because of that human curation signal that we get, that is people styling outfits on our platform or designing rooms on our platform. AI by itself doesn't have style and taste. Humans have style and taste, and then the AI can learn from that. So when humans come to our platform and they design things, put together outfits and say which handbag looks good with which dress or which sofa would look good in a room setting, that's human taste and curation. Then we can train AI from that, not only to make better recommendations to that user, but to make better recommendations to other users. So the next person comes in, starts with just that handbag, but says, 'You know what? I really like that handbag, but I'm not sure how to style an outfit around that.' Well, we see the intersection not only of how other people style that handbag, but other people with a taste similar to yours styled that handbag. So that's what's letting us do really unique things with AI. And back to our latest multimodal visual search models outperforming by 30 full percentage points the large proprietary models, it's that curation signal plus our compact fit-for-purpose models. As we think forward, we think we're just getting started in terms of what a true AI powered shopping assistant can be. Right now, we're making great recommendations of things you want to buy, helping you go take action on those things. How much more helpful can that be? That is a moonshot in and of itself. Can that get to a place where it's as good as the person you'd love going shopping with the most? Whether that's a best friend, a sister, a brother, or if you're lucky enough to have a personal stylist, can it be as helpful as that? We think that in and of itself is a moonshot and doesn't require a different form factor or things like that. And the last thing I'd say, so many moonshots end up either not panning out or come to fruition much later. I've built five startups from zero, Venmo and Braintree being the two most recent. I've been in Silicon Valley for a long time. And one of the things that I've learned is that the best moonshots are often the ones that are deeply grounded in a real human need and a unique signal, not just a shiny new technology.

Two things you'll hear in Silicon Valley. One is like, well, I was right but early. The other thing that any good VC would say in response is right but early is the same as wrong. And so with these moonshots, you've got to have a vision for where you want to go, but you've also got to be deeply tuned into what users are ready for today and what you can deliver today. I think we're balancing that really well. We have a view of like, okay, shopping and human curation are at the core of our platform. That human curation is a real differentiator for us. We think over time that can let us do tremendous things to be as helpful as going shopping with your best friend or the really great sales associate at the boutique. We think that's where that can go over time. What are users ready for today? Today, it's really great recommendations, really seamless ability to go purchase right inside of our platform. You can link an Amazon account and purchase right inside of our platform. Those kinds of things that really make their journeys easy today.

Yeah. As much as we talk about agentic in Silicon Valley, the average consumer isn't thinking about agentic at all. They're just thinking about did this thing help me do what I wanted to do. And on that, I think we're getting really high marks as evidenced by users flocking to the platform and engaging deeper and deeper over the last three years.

Oh, thank you so much, Akos. Really appreciate it. That was Bill Ready, the CEO of Pinterest here on TITV.