Nvidia has been doubling down on the opportunity to build robotics and other industrial AI applications, with the launch of its Omniverse platform, and most recently Mega, an Omniverse Blueprint framework to create digital twins to operate these applications. It’s also investing in digital twin startups to get the effort off the ground.
Taiwan’s MetAI has developed a model that can quickly generate “SimReady” (simulation-ready) digital twins using AI and 3D technology, by converting CAD files into functional 3D environments within minutes.
Now Nvidia is backing MetAI in its first round of funding, a $4 million seed round that becomes the chip giant’s first investment into a Taiwanese startup. Others in the round are a mix of other strategic and financial investors, including Kenmec Mechanical Engineering, Solomo Technology, SparkLabs Taiwan, Addin Ventures, and Upstream Ventures.
The next wave of AI, known as generative physical AI, relies on physically accurate simulated environments to train and validate robots used in autonomous systems, to build operational AI before deployment. MetAI contends that the digital twins that it helps create will be central to that effort.
“Digital twins have long been seen as a barrier to entry for physical AI due to the months or even years of effort required for development,” said Daniel Yu, CEO and co-founder of MetAI, in an interview.
MetAI focuses on AI-powered digital twins tailored to advanced semiconductor fabs, smart warehouses and automation. It also generates synthetic data within AI-enabled digital twin environments.
Renton Hsu, Yu’s co-founder and the CTO of MetAI, has a background in 3D engineering and AI, and he first got to working with digital twins while building enterprise AI software applications: they were used as a practical workaround in situations where clients lacked enough data to train their systems. He then realised he could apply the same to 3D systems, integrating 3D technology with AI to develop synthetic AI and 3D solutions, joining forces with Yu (who comes to the startup with experience in digital transformation projects) and a third co-founder, Dave Liu (COO), to start MetAI.
The breakthrough was enough to win first place in a a competition run by Nvidia, making Hsu a “Jetson AI ambassador” for the country.
MetAI’s competitors are the range of large and small companies that have built digital twin technologies for manufacturing. They include Siemens Digital Industries, Dassault Systemes, Hexagon AB, Duality AI, and Intagles. In the synthetic data sector, there are scores of companies, including Sky Engine and Scale AI.
MetAI believes that it has a unique approach compared to all of these.
“Unlike competitors that prioritize operational efficiencies or IoT integrations, MetAI leverages generative models and AI-driven layouts to create digital twins designed for physical AI training and implementation in real-world operations,” Yu said. “This approach not only accelerates the creation of digital twins but also ensures their direct usability for advanced automation systems like robotics, bridging the gap between simulation and reality.”
MetAI differentiates itself by producing artificial data within its AI-enabled digital twin environments. Yu noted that it enables users to generate synthetic data customized for specific operational requirements, facilitating AI training and validation. “Instead of creating isolated datasets, MetAI builds dynamic virtual worlds (i.e., world simulators) — realistic virtual environments that operate exactly like the real world,” he said.
The two-year-old startup — whose products range from vertical AI agents through to digital twins — has a handful of customers and is already generating revenues by partnering with enterprises in the manufacturing and automation industries, and this year it expects to bring in $3 million from a single project, Yu said. Revenue is derived from project-based income, product subscriptions, and licensing fees from ongoing developments, he added.
“MetAI’s integration with NVIDIA Omniverse represents a transformative step forward for industrial digital twins and physical AI in simulations,” Nico Caprez, corporate development manager at Nvidia, said in a statement. “Their ability to create scalable environments for AI training will potentially set a new standard for industries ranging from manufacturing to robotics.”
In 2023, MetAI collaborated with Kenmec to create digital twins for automated warehouses. MetAI’s technology claims to have drastically reduced the time needed for warehouse digital twin simulations from thousands of hours to just 3 minutes, resulting in significant savings in operational and verification tasks.
With the latest funding, MetAI plans to expand its R&D team for faster development and execution of its go-to-market strategies to meet growing demand. On top of that, the Taiwan-based startup intends to establish a U.S. office and relocate its headquarters in the second half of 2025, Yu told TechCrunch.
“Taiwan serves as our testing ground, where we collaborate with Taiwanese industry leaders to integrate deep vertical knowledge into our models, ensuring our solutions are both robust and scalable,” Liu said. “Given its size and the demand for simulation-driven solutions, we are expanding the U.S. market due to high labor costs and operational complexities. Our expansion strategy focuses on providing both point solutions and end-to-end solutions, including SaaS offerings and vertical AI agents designed for rapid implementation in real-world scenarios within these industries.”
