Decart just raised $300 million to rewrite AI’s real-time playbook. The company builds AI world models that work at lightning speed, targeting industries where lag kills the user experience.

Unlike traditional AI that processes tasks offline—like generating images or answering questions—Decart’s tech thrives on instant interaction. Its software runs on chips from Nvidia, Google, and Amazon’s own Trainium hardware. The goal: power AI that reacts in milliseconds, not seconds.

Decart’s flagship platform, the Decart Optimization Stack (DOS), compresses model training and inference workflows from months to weeks. It boosts token processing rates eightfold over industry averages, hitting 1,600 tokens per second. On video, DOS can handle full HD streams at 100 frames per second—an order of magnitude faster than most AI video models.

This speed fuels two core products: Lucy and Oasis. Lucy manipulates live video streams in real-time, enabling virtual try-ons and interactive advertising. Oasis tackles physical AI, targeting robotics and autonomous systems with real-time environmental simulations. Both rely on Decart’s optimized stack to deliver seamless, lag-free experiences.

Amazon’s investment and partnership stand out. Lucy2 runs on Amazon’s Trainium 3 chips with over 80% chip utilization—a rare feat signaling efficient hardware use. For a company less than three years old, Decart’s $450 million total funding and near $4 billion valuation show investor faith in its real-time AI thesis.

Decart’s founders argue that language models operating purely on text fail to grasp the physical world’s complexity. Their solution: vertically integrated optimization across hardware and software layers, not just bigger models. This approach aims to bridge the gap between static AI outputs and dynamic, interactive AI applications.

The company’s investor roster reads like a who’s who of tech and entertainment: Nvidia, Sequoia, Benchmark, Adobe, Toyota, plus angels like Andrej Karpathy and Michael Eisner. This mix reflects Decart’s ambitions across gaming, media, and infrastructure sectors.

Real-time AI remains a bottleneck in industries craving instant feedback—think live gaming, immersive shopping, or autonomous vehicles. Decart’s software promises to unlock these markets by delivering AI that keeps pace with real-world demands.

While Decart is tight-lipped on exact revenues or client names beyond Amazon, its upcoming releases—DOS 2.0, Lucy 2.5, and Oasis 3—will test whether this vision can scale commercially.

Decart’s rise highlights a broader trend: AI infrastructure must evolve beyond offline batch processing. The future belongs to companies that build AI systems capable of real-time decision-making and interaction, running efficiently on diverse hardware platforms.