Decart, a young AI company, just raised $300 million to build real-time world models. Their goal is to bring AI closer to how humans experience the world. This fresh funding pushes their total to over $450 million since they started in 2023.

What makes Decart stand out is their focus on speed and scale. Their platform, called the Decart Optimization Stack (DOS), runs AI models that process data much faster than usual. For example, it can handle 1,600 tokens per second, compared to the typical 200 tokens most systems manage.

They run these models on major AI chips like Nvidia GPUs, Google TPUs, and Amazon’s Trainium. Their partnership with Amazon is a big deal because Decart’s “Lucy2” model runs on Amazon’s Trainium3 chips and uses over 80% of the chip’s power efficiently. That means more of the hardware’s potential goes directly into AI tasks.

The Power Behind Real-Time World Models

Decart’s technology centers on what they call “world models.” Unlike traditional language models that only understand text, these models grasp how the physical world works in real time. This matters a lot for AI systems that need to interact with the real environment, like robots or virtual experiences.

Lucy, one of their flagship products, is a world model designed for immersive experiences. It reacts to user input in under 30 milliseconds. Retailers use Lucy for virtual try-ons and dynamic video ads that change based on what the viewer does or says. This makes online shopping more interactive and personal.

Another product, Oasis, targets the physical side of AI. It supports robotics and autonomous systems, building on a viral demo that showed Minecraft-style real-time AI environments. This means robots could better understand and navigate the physical world using Decart’s AI.

Backing from Big Names and What’s Next

The company’s investors include Radical Ventures, Nvidia, Sequoia, Benchmark, Adobe, and Toyota. Even high-profile angels like Andrej Karpathy, who helped build Tesla’s AI, and former Disney CEO Michael Eisner are on board.

This mix of investors shows Decart’s broad appeal across media, gaming, infrastructure, and robotics. Their technology fits many fields where real-time AI and video processing are crucial.

Decart plans to release updated versions of their products soon. DOS 2.0, Lucy 2.5, and Oasis 3 are expected in the coming weeks. These upgrades should improve speed and accuracy, pushing the limits of what AI can do live.

Decart’s story is also an example of how AI startups work with chip makers. Nvidia has invested heavily in AI companies this year, often tying equity stakes to long-term hardware commitments. Decart runs on multiple chip types but remains a key Nvidia partner.

Real-time AI is the next big challenge for the industry. Most existing video and world models lag behind when it comes to speed and scale. Decart aims to fix this bottleneck and unlock new possibilities in gaming, shopping, streaming, and robotics.

The company hasn’t shared its revenue or valuation yet but is clearly building toward a future where AI models understand and react to the world as fast as we do. That’s no small feat, but Decart’s progress so far shows it’s within reach.