SambaNova CEO: AI’s Next Phase Is About Efficiency

This week, you are closing your series f fundraising round for viewers who paying attention. You already had kind of an announcement of a closing in July. This one is different, though. We were just talking about you don't just have financial backers. You have strategic corporate backers. Talk to me about the significance of bringing in companies into the fold beyond just financial firms. Yeah. No. It's it's so exciting to see. We've got so much interest, and we did that first close back in July when when when you and I talked. And so all this interest, and then we had to extend the close just because these start coming in. These are huge tech companies who are coming in and partnering with us to actually help someone someone else grow in the market. So they're partnering with you. Does that also mean we're gonna start to see large tech companies double down and use SambaNova technology too? Yes. So, yeah, I mean, you're you're seeing this incredible interest in the market with with inference growing, and people are looking at heterogeneous environments using chips from NVIDIA and AMD and SambaNova to intermix into... In in order to drive the best economics. And so we're seeing these these data centers come up. They're all gonna have all these hybrid hybrid environments, and we're super excited to be part of that. Say, another reason this conversation is so apt for today is we had JPMorgan earnings. And on the earnings call, Jamie Diamond was talking about this idea. We are using a wide array of chips. They are one of your customers too. What what is JPMorgan using the SambaNova ecosystem for exactly? Yeah. We're really excited with the partnership. We have r racks at JPMorgan behind their firewalls, and this is one of the things that's really important for enterprises where security, safety is important. People are bringing their technology back on prem, bringing behind their own firewalls, protecting their customers' data behind those firewalls. And because some of it was so power efficient, so low cost, we're able to deliver that on their own premise inside their own data centers. Does that also necessarily mean if they're using your technology on prem, maybe using their own model or open source, that companies like JPMorgan are gonna start drifting away from using OpenAI, from using Claude, from using the large language models? Well, I think it's gonna be hybrid. I think these frontier models are just tremendous, and they do all sorts of great things. And, actually, some of them who work closely with those frontier labs too and so that you can intermix. There's some things where those models are just really, really good at and some things where companies wanna keep private, companies wanna keep secure. They have their own IP, and so they create their own models. They run it privately and security on Samanova on prem, and they don't have to worry about that information getting out. Well, part part of the reason I ask is the Broadcom CEO was also speaking on Monday at a conference. He had a lot of questions about the ecosystem, but about these large language models saying they're losing money and questioned investors into them. And I wonder if you also question the economics of it because your very existence shows that they're not the only game in town. Yeah. Well, look. You know Broadcom's a great partner of ours, and, and we've been talking about it. In the in the world of inference, economics is paramount. Right? That as you scale, it's about how quickly can you actually deliver returns on that capital investment you're making. And so with Summit Nova, we're able to give you six month returns, where, otherwise, it might take you eighteen, twenty four months to do it. And so as we're growing, I think you're gonna see people be much, much more conscious about dollars, return on investment, how how quickly can you get your payback on the investments you're making. Rodrigo, again, all of that kinda feels like the future that we see for these large language models and thus the build out of things like NVIDIA and things that are training models. Do we go too far then? Is there a risk that we are going too far if the world is waking up to things that SambaNova offers that are cheaper and are more secure in a closed ecosystem? Well, I think we're in early stages still despite all this build out. You do look at most enterprises are just getting started with delivering applications on AI, and so I think we're just at the early days. Now we're entering the phase where it is about efficiency. It is about... You know, you think about data centers, and some of them will focus so much on energy efficiency so that we can actually reduce the number of data centers required. And so we wanna focus on making sure that we deliver technology that can scale and allow people to use it without all this extra cost that today we're we're facing. I I was also thinking about, Haktan's comments sort of ironically coinciding with this close you're doing because one of the things he said was one of the biggest bottlenecks was financing. And wonder what you think of that comment considering you've been able to bring funds in. And I wonder if you think perhaps other parts of the ecosystem are a little bit harder to finance now. Well, look. I mean, it's a capital intensive business. And as you scale, you're having to put all these chips into the factory and taking nine to twelve months in order for you to actually stay in the data centers. And so I think all of that's going to be basically capital that's in process of delivering value. Right? And so I think we're gonna need to figure out how to actually accelerate that because the growth is so extensive, and and so we need to actually make sure the capital is there to support it. I I would love to talk about your growth because, again, the number that you had in July was 11,000,000,000. We'll wait for a final announcement to see where that goes. So at that size, you're about 4% of arms market cap, 1% of AMD, just a rounding error error to NVIDIA. Do you see a future where your size rivals those? Oh, look. I think it's an exciting time to be... You see these companies where a year before that, it was a quarter the size maybe. And so the growth is incredible. What we have to do as a company... And look. We're we're a start up still, and we're focusing on growing, but we're actually seeing this market turn right into what we're good, which is inference. It's about energy efficiency. It's about cost efficiency. And so if you are are thinking about that, SunBanova becomes a very, very good option as you scale your inference workloads. Well, and presumably, you have less of the whole data center backlash. How are you viewing that and how you position the company considering some of these other large data center projects have stalled with the regulation and the local pushback? Well, we're partnering with Vista on this. We're super excited about the Neo Cloud that Vista is building VC two. And what what their strategy is and some of is using existing data centers, reusing it. You know, taking what Neil called brownfield data centers that already have the space, already have the power, already have the infrastructure developed, and actually reusing it and reenergizing that. It's all over the world in terms of how many of those exist. We just gotta fit it into that power footprint. And so with Aminova at 10 kilowatts versus a 100 kilowatts per rack, we're able to actually use... Reuse all the data center, which is actually gonna help not just, businesses, but also the environment. How receptive are are people to that message? And part of the reason I asked that is a lot has been said that part of the issue with AI and the data center pushback has just been... They haven't done a good job marketing set... Themselves, that they've it's been doom and gloom, and the safety is a concern and not things like, hey. Maybe this stuff can cure cancer. And I wonder as as a business leader in this space, what you see is your responsibility. Well, incredibly important for these things to come out. And if you kinda look back twenty years ago in the Internet, all these things were coming up around security and privacy, and then the entire industry around cybersecurity was born out of that. And this is great that these conversations are happening now because safety is important. Right? Efficiency is important. It's about making sure that we actually have sustainable technologies. And so we're all part of this. We're doing our part. We're driving the power down. We're driving the cost down, making sure that we're reusing data center... Centers that already exist. Right? And so we all have to be more conscious about how we actually scale this because it's gonna be important. Do you think that that's the ultimate bottleneck? Maybe not capital, but the pushback you're getting in local communities for others? Well, I think it's important. Right? Sustainability is important, and I think you're gonna have to figure out where these go. And so we start by saying, look. Can we deliver the value and the technology you need by reducing the power that's required? And so that that... If you start there, it actually opens up opportunities in a lot of different Rodrigo, tracking the Sampanhova's journey, just how it exists as a company has been really interesting. There were rumors that you and I talked about, you know, months ago about Intel partnership. You've addressed whether you would go public before. You've said maybe if it makes sense for you, but now you have the capital, there's no need for that. When would that change? What would need to go right or maybe even wrong, but what would need to go right in order for you to consider becoming a public company? Well, look. I think we're in this high growth mode right now, and so we're... We are partnered with Intel. We're partnered with several others because, you know, as a company in this space, partnership is really important, and you wanna continue to actual, you know, leverage, you know, that into into the business because, you know, it's not just one piece of technology that you need. You need all the, CPUs and the memory and the storage, and so you need to partner with everybody. And so as we do that and we grow to the next level, we can see what the capital needs are. And you see even the largest companies, you'll be looking for, additional capital. So there are moments where we're gonna see, do we need more capital in order to actually take us to the next level? I see. So it basically becomes a matter of you need that extra juice that public markets would give you. But for now, in your growth stage, you're not there yet. Is that is that kind of the right No. We're just focusing on execution right now. I mean, right now, it's about, you know, delivering value, getting the customer wins, getting someone over technology deployed. Supply chain's a big issue, and so we're always kinda making sure that our customers get the supplies they need. And so so we gotta continue to execute and create that flywheel. And at some point, if it makes sense, we discuss it.