Decart Raises $300M to Build the Future of Realtime AI

追加: 2026-05-25

[00:00:01]Okay, huge news coming out of the cart.

[00:00:03]I'm excited to share that we've raised $300 million to scale the next generation of real-time AI. Both through our breakthroughs in real-time world models and our ultra optimized infrastructure powering mode. Today we have news for you with three technical breakthroughs that the world has never seen before. DOSs, Lucy, and Oasis. We built the best infrastructure for low latency AI. It's called DOSS, the cart optimization stack that lets you inference and train models for low latency, an order of magnitude faster than any other system out there.

[00:00:36]Basically, we're making AI feel instant.

[00:00:39]AI becomes alive. With DOS 2.0, we can run Lucy Oasis at full HD at 100 frames per second. We're very thankful to have partners across the industry like Amazon, Nvidia, Adobe, eBay, Toyota, and others joining our journey.

[00:01:03]Dean, I'm so excited being here. Took us way too long to partner up. I think I've known you now for 2 years, 2 and a half years or so. And as you know, my initial fascination with this topic of world models came from looking at it through video games. The future of interactive media and obviously AI is playing a huge role. And I think over the last year or two, the narrative on world models has actually changed and evolved quite a bit. Now we understand it might actually be a key ingredient for this quest toward AGI. I think that world models are going to be a key piece of AGI. AGI is not going to be just just this one thing that happens. It's going to be a system combining multiple different aspects to it. AI has always had three pillars. It had NLP which is text. It had computer vision which is which is structure which is pixels and RL and reinforcement learning. And now we're seeing that all three of these components are coming together to enable the next frontiers of AI that will ultimately lead us towards AGI.

[00:01:56]>> And a big part of this is your Decard optimization stack or DOS now claiming two orders of magnitude speed improvements over other solutions. How does that work? So actually our first product at Dart even before the world models was DOSS the cart optimization stack and and we built it to be the most efficient inference and training engine across all three different hardware GPUs TPUs and tranium for a variety of AI models. It turns out that to get the best performance out of AI hardware and to build the best AI models you can't separate the two. You have to understand both. You have to understand both how to build models, how to write the software and how the hardware circuits themselves are built. And only by understanding all three parts can you really get for example to 80% MFU and to get to the highest performance possible on these chips.

[00:02:42]>> Well Dean this is quite the round and coalition you're bringing together here.

[00:02:46]You have Nvidia and eBay, Adobe, Toyota.

[00:02:49]You have the Nintendo family. You have the former Disney CEO Michael Eisner and many others joining. What does that tell us about the future of the card? Where is this headed? Well, we're very fortunate to have the entire economy across different industries, different markets join us in in taking the tech that we're building at the cart and bring it to market across different opportunities in gaming, in media and interactive experiences, e-commerce, live cryion, live streaming, and all the way to physical AI, robotics, manufacturing, logistics. We're creating tech here that is very foundational for AI. AI is going to uproot and change many different industries and we have lots of different potential opportunities for what we can do with the tech at the cart and many that we're actually already deploying in production >> and for a lot of these names we just mentioned you're already generating significant revenue today.

[00:03:38]>> Well yes I think that both DOSs which is what we started productizing first and is now full stream with DOS 2.0 is generating much value for customers in creating much more efficient AI across chips software and models. This comes in parallel with the models Lucy taking off in e-commerce and live streaming and actually generating new value for customers across these industries at the application layer as well.

[00:04:01]>> We're glad you brought up Lucy because I would love to talk a little bit more about Lucy and Oasis.

[00:04:05]>> Lucy Lucy and Oasis are are the crown jewels. They're they're the models that pioneered world models and now are at the forefront of this new category. It's already taken off in e-commerce and live streaming and we're seeing early adoption in gaming and in new ad formats. You know, with Lucy, we're creating new experiences online. We ask that you change yourself live into anything.

[00:04:26]>> Like anything, >> anything, >> anything, >> anything, anything, anything, anything.

[00:04:30]You can really do anything.

[00:04:32]>> I think I got it.

[00:04:33]>> Okay. Okay. Okay. Enough with the questions for me.

[00:04:36]>> Yeah.

[00:04:37]>> Maritz, why are you investing?

[00:04:38]>> I mean, for me, I'm coming at it very much with a gaming and interactive media lens. And I believe with AI, a lot of the focus is on just building existing things faster, cheaper, and the new unicorns and decacorns that are being born. They are the companies that build previously impossible experiences. And I think this is exactly what world models and what the card is about. That gets me excited and I hope we can work on some interesting stuff together.

[00:05:06]>> So Ritz, how do you how do you buy your clothes today? Cuz cuz this is style.

[00:05:10]>> Oh, thank you. Thank you. I don't love shopping, honestly. I do most of it online and like I feel like I return half of the stuff I buy.

[00:05:17]>> So So I think I think you're doing it all wrong. There's a completely new version.

[00:05:21]>> Okay.

[00:05:21]>> So I think you know if if it be a bit a bit colder, >> you should put on something like this perhaps.

[00:05:27]>> Oh, look at that.

[00:05:28]>> Okay. You can you can even interact with it.

[00:05:31]>> This is crazy.

[00:05:32]>> Sometimes sometimes you even want to try out completely new outfits. Changing the clothing itself, the the shirts, the pants, everything. You can take a look.

[00:05:41]You can turn around. You can see how it looks on you. Can I get a coffee? Any shots?

[00:05:46]>> Wait, wait, wait a sec, guys. Wait a sec. See, see this water? Yeah. See this water?

[00:05:51]>> Oh, boy.

[00:05:53]That works great.

[00:05:54]>> Imagine if instead of replacing this product as you're as you're here, you can do it when you're live streaming.

[00:05:59]And you can also do it not just in the real world, but you can do it when you look like this.

[00:06:06]>> Well, there's really almost no delay at all. It's just literally instantaneous and we can become pirates right here, right now.

[00:06:12]>> This is a completely new lifestyle.

[00:06:15]I love this one. Nice.

[00:06:19]I'm excited for what's next. This will be the quickfire segment, or as I like to call it, the low latency segment. For that, I'll read you some questions from the cards.

[00:06:29]First thing you do when you wake up in the morning, >> brush my teeth.

[00:06:32]>> And we're very thankful for that.

[00:06:33]>> What drives you?

[00:06:34]>> We have to build AGI. We have to build it now. We have to completely change the world with it. Tell me about when was the last time where you changed your mind on something.

[00:06:41]>> Every single day.

[00:06:42]>> Will robots replace the co role too?

[00:06:44]>> Yes.

[00:06:44]>> Perfect. When?

[00:06:45]>> Now.

[00:06:46]>> What's your favorite way to clear your head?

[00:06:47]>> Going on walks.

[00:06:48]>> What is something you refuse to compromise on?

[00:06:50]>> Speed. We have to move as fast as we possibly can. That's the only mode in AI.

[00:06:54]>> What would happen to you if your teammates cut your hair?

[00:06:56]>> That would be very, very sad. And all our power would drain away.

[00:06:59]>> What's one word to describe this current era?

[00:07:01]>> Insane.

[00:07:01]>> What's your ultimate comfort food?

[00:07:03]>> Pizza.

[00:07:03]>> What's your favorite part of the office?

[00:07:05]the bean bag room that we get to sleep in and work in at the same time.

[00:07:08]>> In one sentence, what is the card building?

[00:07:10]>> We're building the foundations that allow AI at insane speeds to come here as fast as possible and as smart as possible.

[00:07:16]>> Why now?

[00:07:17]>> It's all happening right now. It's happening way too early. AGI is 20 years too early. We got to do it now cuz that's what's happening.

[00:07:22]>> And why you dream since day one since since I was 8 years old.