Decart just closed a $300 million funding round. The company now sits on over $450 million in total capital. Investors include Radical Ventures, Nvidia, Sequoia, Toyota Ventures, and notable angels like OpenAI’s Andrej Karpathy and former Disney CEO Michael Eisner.

Decart builds AI infrastructure designed for real-time, low-latency applications. It offers three main products: the Decart Optimization Stack (DOS), Lucy, and Oasis. DOS accelerates AI training and inference across hardware platforms. Lucy and Oasis are world models—software that simulates dynamic virtual and physical environments in real time.

The company claims DOS 2.0 processes AI workloads eight times faster than industry norms. It can handle over 1,600 tokens per second and 100 frames of HD video per second. Lucy modifies live video streams instantly, enabling applications like virtual try-ons and in-video advertising. Oasis simulates physical spaces with real-time feedback, useful in robotics and autonomous systems.

Decart’s technology integrates with major hardware, including Nvidia GPUs, Google TPUs, and Amazon’s Trainium chips. The partnership with Amazon Web Services is strategic; Decart runs its models optimized for AWS’s custom silicon. This collaboration aims to scale real-time AI for enterprise clients globally.

CEO Dean Leitersdorf frames world models as the next AI frontier. Unlike language models that only process text, these models simulate environments that persist and react physically. This shift enables AI to interact with the real world—powering robotics, immersive gaming, and new commerce experiences.

Decart’s revenue reportedly grows from cloud contracts, AI labs, and large tech firms licensing DOS. The company emphasizes cutting the time and cost to optimize AI models for different chips. Instead of months, DOS reduces deployment preparation to weeks. This flexibility supports shifting workloads across hardware with minimal downtime.

Despite the hype, real-time AI remains a tough problem. Current video generation lags by 100 times the speed needed for smooth, interactive experiences. Decart’s approach rebuilds AI infrastructure from the ground up, focusing on hardware-level efficiency and specialized world models.

The $4 billion valuation reflects investor confidence in Decart’s vision and execution. It also highlights the race to build AI systems that operate at human timescales. Companies betting on static, offline AI risk falling behind as demand for interactive, real-time applications explodes.

Decart’s growth parallels a broader trend. Firms like Armada push modular AI data centers to bring compute closer to data sources. Meanwhile, Decart focuses on the software stack that squeezes maximum performance from cutting-edge chips.

In a landscape crowded with AI startups, Decart’s deep integration with hardware partners and ambitious world models set it apart. Whether it can deliver on the promise of seamless real-time AI remains to be seen. But the funding and strategic ties ensure it will shape the conversation for years.