Decart Lands $300M to Advance Real-Time AI

The big picture: Decart, an AI research lab specializing in real-time video and world models, announced on Monday it had secured $300M in new funding led by Radical Ventures. This latest round brings the two-year-old company’s total capital raised to over $450M.

Why it matters:

• Strategic Investor Mix: The funding round attracted a diverse group of investors including Nvidia, Sequoia, Benchmark, Adobe, and Toyota, alongside angel investors like OpenAI co-founder Andrej Karpathy and former Disney CEO Michael Eisner. This mix signals Decart’s positioning across media, gaming, and infrastructure.
• Real-time AI Bottleneck: The industry faces a significant challenge in achieving real-time, low-latency inference at production scale for AI-generated media and immersive experiences. Decart’s focus on this bottleneck addresses a critical market need as AI adoption expands beyond text-based models.
• Multi-Cloud Strategy: Decart’s optimization stack (DOS) runs across Nvidia GPUs, Google TPUs, and Amazon Trainium, positioning it as a key player in a multi-cloud AI infrastructure landscape. This broad compatibility offers flexibility and resilience for customers.

How it works:

• Decart Optimization Stack (DOS): DOS is an inference and training platform designed for high performance, delivering 1,600 tokens per second for agentic inference and full-HD video inference at up to 100 frames per second, significantly exceeding industry averages.
• Lucy World Model: Lucy is Decart’s ‘world model for immersive experiences,’ capable of responding to user input in under 30 milliseconds. It is currently deployed in virtual try-on, live streaming, and dynamic in-video advertising applications.
• Oasis for Physical AI: Oasis is Decart’s parallel product for physical AI, targeting robotics and autonomous systems. Its capabilities were showcased in a viral real-time Minecraft-style demo, demonstrating its potential for understanding and interacting with the physical world.

The catch: While Nvidia is a significant investor, Decart’s claim that its DOS stack runs across all three major chip families (Nvidia, Google, Amazon) presents a nuanced dynamic. This multi-platform approach tests Nvidia’s typical equity-for-GPU commitment model, as Decart is both a portfolio asset and a public reference for competing hardware. The company did not disclose its current valuation or run-rate revenue, making it difficult to assess its financial trajectory relative to its substantial funding.