Decart, an Israeli AI startup, has raised US$153 million in funding since its founding less than two years ago, including a US$100 million round at a US$3.1 billion valuation completed on August 7, 2025.

The company develops real-time video generation models and claims its technology can run at lower costs compared to competitors such as Google and OpenAI.

Decart reports that its flagship product, Oasis, reached 1 million users within three days of launch, while its follow-up model, Mirage, enables live video transformation using text prompts.

The firm has expanded to more than 60 employees and is establishing a research and development center in San Francisco, led by Dr. Kfir Aberman, formerly of Snap and Google.

Decart says it has spent less than US$10 million of its total funding so far, with revenue from GPU acceleration and video licensing covering operating costs.

🔗 Source: Calcalist

🧠 Food for thought

1️⃣ Israel’s military intelligence units create a unique talent pipeline for AI startups

Decart’s founding story reflects a broader pattern in Israeli tech where military service in elite intelligence units translates directly into startup success.

The Israel Defense Forces, particularly Unit 8200, plays a crucial role in nurturing future tech leaders through hands-on experience in advanced technologies1. Military service in these elite units equips many Israelis with skills that translate well into the startup ecosystem2.

This military-to-tech pipeline helps explain why Israel ranks as the third highest country globally for startup activity2, despite its small size.

The combination of technical training, problem-solving under pressure, and network effects from serving together creates a unique advantage for Israeli entrepreneurs entering competitive markets like AI.

Unit 8200 veterans have founded numerous successful tech companies, suggesting that Decart’s founders bring proven methodologies for scaling complex technical projects.

2️⃣ Real-time video generation solves fundamental computational bottlenecks that have limited AI applications

Decart’s achievement with 20 frames per second and sub-100ms delay addresses core technical challenges that have prevented widespread adoption of AI video generation.

Real-time video processing requires algorithms to analyze and generate video frames instantaneously, facing significant bottlenecks in processing power and computational demands3. Key technical hurdles include latency issues and the need for efficient data throughput to handle high-resolution video streams3.