Nvidia has doubled down on opportunities to build robotics and other industrial AI applications, with the launch of its Omniverse platform, and most recently Mega, the Omniverse Blueprint framework for creating digital twins to operate these applications. It also invests in digital twin startups to get the effort off the ground. Taiwan’s MetAI has developed a model that can rapidly produce “SimReady” (simulation-ready) digital twins using AI and 3D technology, turning CAD files into functional 3D environments in minutes. Now Nvidia is backing MetAI in its first round of funding, a $4 million seed round that is 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, Solomon Technology, SparkLabs Taiwan, Addin Ventures, and Upstream Ventures. The next wave of AI, known as generative physical AI, relies on physically accurate simulation environments to train and validate robots used in autonomous systems, to build operational AI before deployment. MetAI insists that the digital twin it helped create will be at the center of the effort. “The digital twin has long been seen as a barrier to entering physical AI because of 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 manufacturing, smart warehouses and automation. It also generates synthetic data in an AI-enabled digital twin environment. Renton Hsu, Yu’s co-founder and CTO of MetAI, has a background in 3D engineering and AI, and he first worked with digital twins when building the company’s AI software application: used as a practical solution in a client-less situation. sufficient data to exercise that system. He then realized that he could apply the same 3D system, combining 3D technology with AI to develop synthetic AI and 3D solutions, joining Yu (who came to the startup with experience in digital transformation projects) and the third co-founder, Dave. Liu (COO), to start MetAI. The breakthrough was enough to win first place in a competition run by Nvidia, making Hsu a “Jetson AI ambassador” for the country. MetAI’s competitors are a number of large and small companies that have developed digital twin technology for manufacturing. These include Siemens Digital Industries, Dassault Systemes, Hexagon AB, Duality AI, and Intagles. In the synthetic data sector, there are many companies, including Sky Engine and Scale AI. MetAI believes it has a unique approach compared to all of this. “Unlike competitors that prioritize operational efficiency or IoT integration, MetaAI uses AI-driven generative modeling and layout 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 direct usability for advanced automation systems like robotics, bridging the gap between simulation and reality.” MetAI differentiates itself by generating artificial data in an AI-enabled digital twin environment. Yu noted that users can generate synthetic data tailored to specific operational requirements, facilitating AI training and validation. “Instead of creating isolated datasets, MetAI builds dynamic virtual worlds (that is, 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 to digital twins — has several customers and has generated revenue by partnering with companies in the manufacturing and automation industries, and this year expects to earn $3 million from a single project, Yu said. Revenue is derived from project-based income, product subscriptions, and license fees from ongoing development, he added. “The integration of MetAI with NVIDIA Omniverse represents a transformative step for industrial digital twin and physical AI in simulation,” said Nico Caprez, enterprise development manager at Nvidia, in a statement. “The ability to create scalable environments for AI training will potentially set new standards for industries ranging from manufacturing to robotics.” In 2023, MetAI collaborates with Kenmec to create a digital twin for automated warehouses. MetAI technology claims to have drastically reduced the time required to simulate warehouse digital twins 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 market strategy development and execution to meet growing demand. In addition, the Taiwan-based startup aims to establish a US office and move its headquarters in the second half of 2025, Yu told TechCrunch. “Taiwan serves as a testing ground, where we collaborate with Taiwanese industry leaders to integrate deep vertical knowledge into our model, making our solutions robust and scalable,” said Liu. “Due to the size and demand for simulation-driven solutions, we are expanding the US market due to high labor costs and operational complexity. Our expansion strategy is focused on providing point solutions and end-to-end solutions, including SaaS offerings and vertical AI agents designed for rapid implementation in real-world scenarios in the industry.
