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
Decart's valuation hits nearly $4 billion as investors pour capital into startups making AI computing more efficient
Context & Ripple Effects
Decart’s funding trajectory has accelerated from a seed round in October 2024 to a Series A in December 2024 and a Series B at a $3.1B valuation in August 2025. The company’s positioning has also broadened across training at scale, cloud-provider and AI-company tooling, real-time generative video, and GPU optimization.
The latest round matters because it extends that valuation step-up while supplying capital to pursue a technically demanding part of the AI stack: making scarce, expensive compute more productive.
First-order effects
- Decart gains $300M to expand its GPU-optimization and real-time generative-video products, with a valuation near $4B giving it greater capacity to compete for engineering talent and enterprise or cloud-provider deployments.
- Existing investors and new backers are pricing Decart as infrastructure rather than a narrow application company, increasing expectations that its technology can translate into broad AI-workload adoption.
Second-order effects
- Cloud providers and AI companies that use Decart’s tooling gain another potential route to improve the efficiency of training and real-time generation workloads; competing optimization vendors will face pressure to demonstrate comparable performance and integration.
- The financing reinforces investor attention on companies that can lower the compute burden of AI products, not only on companies building new models or end-user applications.
Third-order effects
- If efficiency tooling continues attracting capital at this pace, the AI stack may place more strategic value on software that improves utilization of existing GPUs, alongside continued spending on additional compute capacity.
- That shift could make infrastructure economics—latency, throughput, and cost per workload—a more important basis of competition for generative-AI services, though Decart’s ability to sustain its valuation depends on converting technical claims into durable deployments.
The trend: This is part of the AI industry’s move toward treating compute efficiency and real-time inference as strategic infrastructure markets, rather than merely back-end engineering functions.