Decart, which offers real-time generative video and GPU optimization tech, raised $300M at a ~$4B valuation, up from $3.1B after raising $153M in August 2025
Context & Ripple Effects
Decart’s financing history shows a rapid expansion from an early-stage company helping organizations train AI models at scale into a supplier of real-time generative video and GPU-optimization technology for cloud providers and AI companies. Its reported valuation rose from more than $500M at the Series A stage to $3.1B in the August 2025 Series B.
The company’s related product coverage centers on real-time interaction: Live Stream Diffusion and Mirage, a video-to-video model that can manipulate live footage. The new round supplies capital behind that shift from model-training infrastructure toward live generative-video applications.
First-order effects
- Decart gains $300M to fund deployment and development of real-time generative video and GPU-optimization products, while its higher valuation strengthens its position with cloud-provider and AI-company customers.
- The financing validates investor appetite for Decart’s combination of lower-cost AI compute and low-latency video generation, rather than treating video models solely as offline content-creation tools.
Second-order effects
- Cloud and AI-platform customers have another well-capitalized supplier to evaluate for real-time video workloads, increasing pressure on competing infrastructure and model providers to demonstrate both latency and compute efficiency.
- As tools such as Mirage target live-footage transformation, livestreaming and other interactive-video product teams may face faster expectations for integrating generative effects and real-time AI capabilities.
Third-order effects
- If funding continues to favor companies that pair generative models with inference-efficiency technology, value may concentrate in providers that control both the real-time model experience and the cost of running it.
- The move toward live AI-generated or AI-altered video could make operational questions around reliability, provenance, and platform safeguards more central, particularly where transformations occur during a broadcast rather than before publication.
The trend: Decart is part of a broader shift from training-focused AI infrastructure toward real-time, interactive generative systems whose adoption depends as much on efficient inference as on model quality.