Real money is in usable tools running at scale solving problems today. And nowhere is that shift more visible than inside the enterprise. Companies are watching cost, flexibility, security. They don't necessarily care who tops which model tops the math and coding benchmarks. They care who gets their marketing team to stop filing IT tickets. Which brings us to our next guest. Aaron Levie is the CEO of Box. He sees where this tech works, where it breaks, why open source challengers like DeepSeek are maybe eating into deals once reserved for OpenAI or Google. Aaron, thank you so much for joining.
Yeah, thanks for having me.
Okay, let's start with super intelligence because it's just become this buzzword of the moment. I got to ask you, what does it actually mean in this context? How much do you care about it? How much do your enterprise customers care about it?
I think the fun thing about being in Silicon Valley is we will continuously create new buzzwords on an ongoing basis, and so you can kind of expect that from the industry. That actually makes it kind of fun, and I'm happy that we have a new thing that we're talking about. I think directionally, all that matters is this idea that AI systems and models are becoming more and more intelligent over time and they're capable of executing much more complicated tasks, whether it's coding software, doing very complicated math, being able to do financial analysis, healthcare, drug discovery. So whether you call that super intelligence, AGI, or just the continued vector of improving the capability of these models, it almost doesn't matter. The industry is in a race of really bringing intelligence to the masses at the lowest cost possible, right?
But has there ever been a buzzword that has been worth so much money? I mean super intelligence, it's the difference between OpenAI partnering with Microsoft and becoming a nonprofit. It's really underpinning this insane talent war race. Do you think about it in those terms? Because I feel like Zuckerberg and Altman have more at stake here. Whereas what you're talking about is like using what we have and specializing and making agents better. Do they need to be smarter than humans? Do we even get there?
Yes, we do get there and yes, they need to be smarter than humans, because that gives you the generalizable platform to build on top of. If you think about AI models already in a number of fields exceeding human capability in many different fields, or at least a significant portion of human capability, but if you imagine a world where you have an AI model or many AI models that exceed humans on every dimension, the capability that you would then have inside of an organization or an enterprise or even in our personal lives would be massive. So I think the goal of getting to super intelligence is actually very important. But to me, it just doesn't matter what you call it—AGI, super intelligence, or just continued advancements on the current vector of AI. What is clear is the recognition of how valuable this race is and the importance of each company getting there as quickly as possible. You're now obviously seeing that in Meta's efforts, how much they're willing to pay to catch up to a lot of the leading companies that are out there. This is definitely the most valuable moment in software and technology history. There's never been a period where you have this many companies chasing a singular prize as much as you have with super intelligence or AGI or advanced AI. It's an incredibly exciting time to be in the tech industry and to make it real for the rest of the world. What this will impact is you'll just see continued productivity gains in every single field of work and in every single area or domain of our personal lives as well. That's the great benefit that we'll experience from this.
Right. So there's sort of agents on the enterprise side and the consumer side and this race I think you kind of framed it as a race for super intelligence. That's how I think about it too. But I don't really know what that means. Like how do we know who gets there first? Is it first to declare it? Does it matter if it's Zuckerberg or Altman?
Well, I would be less worried about a binary moment where there's some sort of milestone and somebody kind of calls it super intelligence. It's much more you can see the compounding returns that the earliest players get in this space. The reason that ChatGPT is ahead of everybody else or OpenAI is ahead of everybody else was the first kind of iPhone moment of AI, and so they have a compounding return which is they've become the default brand for so many different people in AI, and you can see the benefit and how that accrues to the companies that get there early. Conversely, if there are significant step function improvements in intelligence in these AI systems, then obviously the companies that get there earliest also will see continued benefits and returns based on that. So it is a race. It does actually matter who gets there first or second or third or fifth, because you get a high degree of a flywheel of more usage, more data, better models, more users, and that is a flywheel that is almost unstoppable for the companies that are able to capture it. This will be one of the most valuable technology races in history, and it already has been.
Right. And I was talking earlier about how it feels like Altman and Zuckerberg, Meta and OpenAI, they're fighting over this talent, trying to get to super intelligence, but what Google seems to be doing—they've already got some of the best models on the market but they're actually deploying. How do you think about that trade-off between theory and collecting a bunch of researchers, the best in the field, versus actually making applications that people are going to use, whether consumer or enterprise, your world?
Yeah. Well, I think everybody's actually doing the deployment as well, so it just shows up in different ways. Inside of the Meta universe, they've already talked about this and it's pretty obvious how this would show up. If you can improve ad targeting or you can improve ad creation for a small business and find them more customers, that's insanely valuable from a productivity gain for the economy and for all of your customers. So in that context, that's already deployment. It's not the specific kind of ChatGPT ask a question and get an answer; it's driving productivity within their customer base. Similarly, ChatGPT obviously has this great bidirectional dynamic: you as a consumer get ChatGPT, which is an assistant you talk to and get answers from, and they also have a research arm that is creating the next set of breakthroughs that eventually will flow into ChatGPT and their APIs. Google similarly has the Gemini products and experiences and then they have DeepMind which is doing advanced research that continues to create breakthroughs. So I think there's no trade-off between deployment and research, and all of the major players are basically doing both simultaneously.
But is there a trade-off especially for capital and attention? I mean you hear that Zuckerberg is going on these personal recruitment missions. What if he was heads down on the product? And also I just want to make this comparison to what the Chinese are doing, right? And this is a little different because they're doing open source which is in many cases cheaper and more practical. I don't know if you read there was a journal piece on how China is eroding America's lead in the global AI race with cheaper alternatives. And what this says to me is that enterprises and consumers, they don't actually care about super intelligence. They care what's available to them right now, the best product.
Yeah. But I again I don't see any tension. Meta was actually one of the first, if not the first, open source and open weights AI provider. So think about Llama as their basically equivalent of deployment. They have Llama, it's probably the most used open weights AI model on the planet—maybe DeepSeek now is competing with that. So I think they actually have a very strong history of deployment.
You don't think DeepSeek has pulled ahead?
I think I don't know all the metrics, but I'm sure on a global dimension it's doing quite well and maybe even pulling ahead. But that's actually why you need the research flywheel. If you don't have the research flywheel, you don't get the next breakthrough. So what Zuckerberg is doing, and now that the team, Alex, Nat, and others, what's so important about that is you need the next set of research to create the next breakthrough that either you open source or that you're able to go and commercialize. There's no way to be heads down in this space from a product standpoint without the research component. So I think what Zuckerberg is doing is necessary, it's important, it's the right move right now. And what's amazing is that we're going to live in a period where there's going to be at a minimum five major powerhouses that are building toward more and more powerful AI systems and AI models, all basically to the benefit of society and to businesses. We're going to get better healthcare, better tutoring, better education, better productivity in the workplace. So it's incredible to have this moment where there's five companies that are all competing with tens of billions of dollars of infrastructure and capex and data sets and training and amazing researchers to basically benefit all of us. There's never been a period where you have this much competition at the platform layer of a technology space. In the operating system world, you had two or three players that were viable, and even then it was much less scale from a spend standpoint than what we're seeing now. In the cloud wars, we had basically three or four major players that emerged and were scaling up. So never before have we had five major players that are all racing to be able to generate this super intelligence.
So at a minimum, and this is just a minimum, I think you could probably expand to probably 10 or more. You have xAI, you have Anthropic, you have OpenAI, you have Google, and now you have Meta. So five major players all with tens of billions of dollars to go deploy against this problem.
I mean Microsoft and Amazon, they're also working on their own models. Do you count them out? And that also leads to another question. Where you've got all of this amazing AI talent going to those five, where does that leave Microsoft and Amazon that don't have that sort of AI statesman?
Yeah. I mean, I think different answers for each one. On Amazon, I think I'm not going to speak for their whole strategy, but I actually think they have a great playbook right now, which is they want to be an infrastructure provider for AI and they want to enable that for really any of the models that will support their infrastructure. So I know that they're working on Nova and other AI models, but I don't get the sense personally that they see themselves as having to win the super intelligence race as opposed to they have to be the infrastructure layer for the super intelligence race. And Microsoft obviously has this interesting dynamic with the OpenAI relationship where they probably cannot do the same amount of diving head first into the space because of the OpenAI dynamic. But I'm sure that they're working on great breakthroughs as well. But if I take your question and expand out my answer, that would mean that there's another three or four players when you look at the hyperscalers and the surrounding enterprise ecosystem or data ecosystem that also is investing in this space and maybe will have a breakthrough. Eight years ago the transformer paper was written and it came out of a group of people and it wasn't known that that would be the overnight thing that would transform the ecosystem. The next research paper might come out of not one of these top AI labs but some other player in the space, and they might then have some vector of success that lasts until that becomes diffuse across the ecosystem. So you never can count anybody out in this industry. That's another fun thing because there could be a new breakthrough that just emerges. Even in this conversation I failed to mention SSI which has Ilya, which is obviously working on the next sort of frontier of what AI might look like. All I know is that we have a constant stream of innovation from an amazing set of talent and players with a lot of capital. So you can't count anybody out at this stage. That's why you're in such a race, because it's up for grabs at the moment.
Right. Just want to remind our audience if you're tuning in, feel free to ask a question. Aaron, so you said there's five. Maybe you could expand that out to 10. I noticed that these are all American names. Would you include DeepSeek potentially in that top five? You would? Okay. And I mean Alibaba's been doing some really interesting things. Xiaomi, ByteDance, it feels like but they're taking a very different mode, right? And they're going all open source. And some of the people I talked to say that this is really an open source arms race and US AI by keeping it closed actually risks losing out on this whole geopolitical question. What is the world's AI going to be built on? Is there going to be five models? Is there going to be one prevailing one? Where do you see that going?
Yeah, I think my bias—tomorrow's 4th of July, maybe I'm just overly biased toward American answers at the moment—but to your point, it's obviously important to expand out. You have Mistral in Europe, you have Alibaba, you have DeepSeek in China. I'm sure there's some amazing teams right now in India working on breakthroughs. So it's important to think about this from a global standpoint as well. Open source, I do think I've been more and more convinced over time that that's going to be a very important dynamic in the space, and probably the dominant models in the future will be open weights in some capacity. It's already been discussed and announced at some point we'll see an open weights model from OpenAI that has been claimed to exceed our expectations. Google already has actually done some great open weights models which are also extremely strong. So it's not just that China is doing open weights. I think US companies have the same kind of instinct and set of efforts at the moment. So it is an open source or open weights war right now, and I'm pretty optimistic and convinced that the American providers will also lean in pretty heavily in this space.
