Palo Alto Studio Connection Silicon Valley and Wall Street. I'm John Furrier here with Dave Vellante, my co-host.
Welcome to the CUBE studio here at the New York Stock Exchange. I'm Jim Allen, co-host of NYC Wired Fintech Exchange, and we talk constantly about AI transforming finance. But there's a much less sexy question underneath all of it: what data are you actually feeding the machine? Because a stock closing at $102 tells you almost nothing about what happened on the way there, who was buying, who pulled the order, how deep was the market, what happened in the milliseconds before that price moved. Joining me now to have a conversation about how that infrastructure is being built is Christina Qi, co-founder and CEO at Databento. Welcome, Christina.
So, help me understand Databento. First of all, I love the name. I mean, it certainly makes sense to me. I was just planning bento boxes for my children this week, but help me understand what it is that you guys are actually building.
So, we strive to be the single source of truth for basically financial information. So, we're a financial market data API provider essentially is what we do in a nutshell. We aggregate financial stock price information from various exchanges or venues across the globe including the NYSE, and then we basically normalize that data and distribute that to currently about 100,000 or so users around the world.
Wow. Okay. So help me understand who is using that data like for what purpose? Are you creating data for quants or for like—I'm guessing probably not retail customers, right? I assume that it's probably quite a technical-to-technical interaction. Bring it to life a little bit, like who is the key customer? Who did you build this with in mind?
So we actually originally built this product for quantitative traders. It could actually be for anybody honestly. So we do have a lot of retail traders who use the product. We have students who use the product for the first time who have never actually touched quantitative—maybe don't even know how to code actually—who can use the product as well. But then also people who are very sophisticated, including some of the largest financial institutions in the world who use the product as well. So it does range very greatly between the different types of users. And I think recently what surprised us is that we do have a new type of user that uses the product: we have a lot of AI labs that have also picked up on the product. So, yeah, I think that's probably been a bigger surprise for us is that we didn't design the product originally for that type of category, but that has been a big tailwind for us in recent times. Yeah.
So, give me a scenario here. You are a quant trader and you want to understand—I'm looking at the board here—Nvidia stock buys between 9:30 and 10 a.m., right? That information isn't necessarily easily ingestible. You guys can actually kind of solve for scenarios like that. Bring it to life, like how does it compare to the world of the Bloomberg terminal, for example, that we're all so familiar with here on Wall Street?
For sure. So the difference and the reason why we started Databento to begin with is actually because I ran a hedge fund before Databento, and when we were dealing with the previous generation of data providers, we noticed that they were designed for human consumption essentially. They were designed around a human basically staring at a terminal product. And we noticed that even today most data is being consumed by machines, not by humans, and so we wanted to design a data provider that was meant for machine consumption, not for human consumption. And that way we could basically help be the data backbone of any kind of product and help other products launch faster, go to market faster. You can build products on top of us. So, think of us as like we're designed with the builder in mind rather than for an end user. If you pull up a stock price on an app today, like any app on your phone, that data might come from us already. So, that's also what we do as well. We do serve a lot of fintech startups that use us to build apps and products on top of us. Yeah.
And if you're, you know, any sort of enthusiast and you pull up data on your phone like the Nvidia stock price, right, you don't necessarily get the underlying data or the story or the trajectory, what brought it there, what that broader story looks like. Is a lot of this around kind of piecing together that trajectory, that broader story, and where is the value in that data? Where is that realized?
Yeah. So what's interesting is we decided to hyperfocus on just what's called market data, which is just the financial stock price data, instead of alternative data, which would be like satellite data on the parking lots or shipment data or earnings calls or things like that that are more broad that might be surrounding that stock price, where it might go or where it could head, stuff like that. And the reason why we decided to focus on just market data is just because, one, everybody needs market data. Every company needs that within our industry in order to basically survive. We've even had a lot of customers who've invested in our company, for example, or want to invest in our company because they're like, "If you die, we die." Market data is a critical need for a lot of these companies, whereas if you go broader to alternative data, it's just less critical of a need for these guys. And so it's harder to sell alternative data and convince companies that they need that type of data. And so it just came purely down to a preference on what do we want to focus on. Plus, my specialty is more in the market data realm. Not saying that alternative data is not important, it's just that that's just what we chose to focus on. Yeah.
So maybe this is—and I'm sure it is—a false assumption, but in this world of AI where we're hearing more and more about data accessibility, about the democratization of data and information, I would have assumed that a lot of large quant houses are building their own pipes, building their own platforms. What is actually happening, though? Because clearly there's a real market need for this. What sorts of conversations are you having with the buyers across those industries?
Yeah. So, we were surprised by the AI labs using us to begin with. I think the reason why they're using our product, one, is that we're a product that AI can't replicate very easily. Because to give you guys a sense of maybe the differences: we're not just simply reselling data. We actually rebuilt the entire stack from the ground up, all the infrastructure behind the data as well. So that comes down to even things like the bare-metal servers, the colocation—all that infrastructure we actually rebuilt from the ground up. And so as a result of that, because it is a lot of hardware behind the scenes, you can't just vibe-code that kind of company from the ground up. And so as a result of that, a lot of these AI labs end up using us as a customer actually, rather than just doing it themselves, just because it's so much easier to use us from the perspective of a customer.
Yeah, that is fascinating, and that is not something I would have guessed when I looked at this company. So, you're not necessarily another cloud license spend on your end or even a frontier lab fund. It's something much deeper. Talk to me through the company. What's the footprint like then? How many of you guys, where do you exist? What does Databento's actual hardware/software footprint look like?
Yeah. So we are colocated currently at various venues globally. We've signed across I think about 15 different venues currently globally, or 50 different physical locations. But in terms of actual venues or so, it's probably going to be about 60 different locations or 60 different venues globally. So hopefully that will give us coverage across US, Europe, and Asia in the nearish future. So that's something that we're really excited about just in terms of our roadmap. Currently we only have US coverage as well as a little bit of European coverage. So what's fascinating is that we do have over 100,000 users on Databento right now, but our product is incomplete, meaning that we only have a limited number of datasets. And the more global coverage we have, the more likely we are to actually gain more customers because a lot of customers need global coverage in order to even switch over to us as their primary data source.
And what is the kind of chicken-and-egg formula for the datasets? Like where is that information aggregated from? How do you collect and build on that?
