Decart Raises $300m to Scale Real-Time AI Infrastructure and World Models

The funding round positions Decart at the centre of the race to build AI systems capable of understanding and interacting with the physical world in real time.

AI infrastructure company Decart has secured $300m in a new funding round led by Radical Ventures, pushing total funding beyond $450m as competition intensifies around low-latency AI systems, world models, and physical AI infrastructure.

The round attracted backing from major technology and venture investors including Amazon, Nvidia, Sequoia Capital, Adobe Ventures, Toyota Ventures, and Benchmark, reinforcing growing investor focus on the infrastructure layer powering next-generation AI applications beyond large language models.

Building the Infrastructure Layer for Physical AI

At the centre of Decart’s platform is its Decart Optimization Stack (DOS), an infrastructure system designed to improve AI inference and training efficiency while reducing compute costs across major hardware ecosystems including Nvidia GPUs, Google TPUs, and Amazon Trainium chips.

Alongside DOS, the company is developing two real-time world models:

• Lucy, focused on immersive and interactive digital experiences such as gaming, streaming, commerce, and advertising

• Oasis, designed for physical AI applications including robotics and autonomous systems
  

Decart argues that world models represent the next major evolution in AI, enabling systems to simulate environments, understand physics, and interact with continuously changing real-world conditions rather than operating purely through text-based reasoning.

Why This Funding Round Matters Now

The investment reflects a broader shift across the AI sector away from standalone language models and toward infrastructure capable of supporting persistent, real-time, environment-aware systems.

As AI companies compete to move into robotics, autonomous systems, live simulation, and immersive digital experiences, demand is increasing for platforms that can process complex physical environments with low latency and high compute efficiency.

Decart’s positioning is particularly notable because it sits across both infrastructure and application layers, allowing the company to monetise optimisation software while simultaneously developing proprietary world models.

The company’s partnership with Amazon Web Services also signals growing hyperscaler interest in AI systems optimised for proprietary chip ecosystems such as Trainium.

A Strategic Bet on World Models and Real-Time Simulation

Decart has already integrated its technology with AWS Trainium hardware, enabling real-time world models to operate at scale inside existing enterprise infrastructure.

The company says Lucy currently powers applications across virtual try-on, in-video advertising, gaming, streaming, and social media with sub-30ms response times, while Oasis is being developed for robotics and simulation-heavy environments requiring continuous spatial understanding.

With DOS 2.0 now launched and updated versions of Lucy and Oasis scheduled for release, Decart is positioning itself as a foundational infrastructure provider for the next phase of AI deployment.

What This Means for the Industry

The Decart funding round highlights several major trends reshaping the AI market:

• AI investment is shifting toward infrastructure and optimisation layers rather than standalone foundation models

• World models are emerging as a key battleground in robotics, simulation, and immersive computing

• Cloud providers are increasingly partnering with AI infrastructure companies to drive adoption of proprietary chips

• Real-time, low-latency AI systems are becoming critical for physical AI and interactive consumer applications

• The deal also reinforces the growing convergence between AI infrastructure, gaming, robotics, cloud computing, and immersive digital experiences.
  

Summary

Decart’s $300m funding round signals growing investor confidence in world models and real-time AI infrastructure as the industry moves beyond text generation and into physical environment simulation.

As competition intensifies around robotics, autonomous systems, and immersive AI applications, infrastructure platforms capable of delivering low-latency, environment-aware intelligence are increasingly becoming one of the most strategically valuable layers in the AI stack.

---

Discover how nuaxia can support your next medical education initiative:

• Find out more about our specialist services - Moore's Outcome Assessments, Educational Needs Assessments and Patient Impact Studies for the Medical Education sector 

• Contact us on: support@nuaxia.com