Decart Raises $300M With NVIDIA Backing for Real-Time World Models on Amazon's AI Chips
Israeli world-model startup Decart closed a $300M Series B at a $4B valuation, with NVIDIA, Radical Ventures, and angels including Andrej Karpathy and Michael Eisner backing its push into real-time interactive video.
Israeli AI startup Decart closed a $300 million Series B at a $4 billion valuation on Monday, with chip giant NVIDIA joining the round alongside lead investor Radical Ventures. The round also drew checks from eBay Ventures, Adobe Ventures, Toyota Ventures, Atreides Management, and Valor Equity Partners, plus existing backers Sequoia Capital, Zeev Ventures, and Benchmark.
Founded in late 2023 by Dean Leitersdorf and Moshe Shalev, Decart is best known for DOS, an optimization layer that squeezes more throughput out of GPUs and other accelerators. The company plans to use the new capital to push three product lines harder: DOS, the real-time interactive video world model Lucy, and Oasis, a world model aimed at physical AI and robotics simulation. Decart says its stack can already drive world models at up to 100 frames per second of high-definition video, fast enough for applications like a smart mirror that swaps clothing on shoppers in real time.
The angel list reads like a who's-who of AI and entertainment power players. Former OpenAI co-founder Andrej Karpathy joined the round, as did former Walt Disney CEO Michael Eisner, members of the Nintendo founding family, and gaming investor Moritz Baier-Lentz. The mix signals where Decart thinks world models matter most: gaming, media, and any interactive surface where latency is the product.
Strategically, the NVIDIA investment is the eye-catcher. Decart has positioned DOS as the software layer that makes world-model inference economical, and it has publicly courted alternative silicon, including Amazon's Trainium and Inferentia chips. NVIDIA backing a company whose entire pitch is portability across accelerators is an unusual hedge, but it underscores how aggressively investors are racing to own the world-models category as video-native AI moves from research demos to production workloads.
Comments
Share your thoughts. Be kind.
Loading comments…
