Nvidia is doubling down on the opportunity to build robotics and other industrial AI applications, with its launch Omniverse platform, and latest Megaan Omniverse Blueprint framework to create digital twins to power these applications. It is also investing in digital twin startups to get the effort off the ground.

in Taiwan YEAR has developed a model that can easily create “SimReady” (simulation-ready) digital twins using AI and 3D technology, by converting CAD files into a functional 3D environment within minutes.

Now Nvidia is supporting MetAI in its first round of funding, a $4 million seed round that becomes the chip giant’s first investment in a Taiwanese startup. The rest of the round is 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 accurate physical simulated environments to train and validate robots used in autonomous systems, to build operational AI before i -deploy. MetAI argues that the digital twins it helps create will be central to that effort.

“The digital twin has 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 adapted to advanced semiconductor fabs, smart warehouses and automation. It also generates synthetic data within an AI-enabled digital twin environment.

Renton Hsu, Yu’s co-founder and MetAI’s CTO, has a background in 3D engineering and AI, and he first worked with digital twins while building AI software applications: using them as a practical which is a solution to situations where clients are lacking. enough data to train their systems. He quickly realized that he could apply the same to 3D systems, integrating 3D AI technology to develop synthetic AI and 3D solutions, joining forces with Yu (who comes to the startup with experience of digital transformation projects) and a third co-founder, Dave. Liu (COO), to start MetAI.

The success 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 range of large and small companies that are building digital twin technologies for manufacturing. These include Siemens Digital Industries, Dassault Systemes, Hexagon AB, Duality AI, and Intagles. In the synthetic data sector, there are several companies, including Sky Engine and Scale AI.

MetAI believes it has a unique approach compared to all of these.

“Unlike competitors who prioritize operational efficiencies or IoT integrations, MetaAI uses generative models and AI-driven layouts to create digital twins designed for physical AI training and execution of real-world operations,” said Yu. “This approach not only facilitates the creation of digital twins but also ensures their direct use for advanced automation systems such as robotics, bridging the gap between simulation and reality.”

MetAI differentiates itself by creating artificial data within an AI-enabled digital twin environment. Yu noted that this allows users to create synthetic data tailored for specific operational requirements, facilitating AI training and validation. “Instead of creating isolated datasets, MetAI builds dynamic virtual worlds (ie, world simulators) – realistic virtual environments that behave 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 already generated revenues through partnerships with businesses in the manufacturing and automation industries, and this year is expected to bring in $3 million from a project, said Yu. Income from project-based income, product subscriptions, and licensing fees from ongoing developments, he added.

“The integration of MetAI with NVIDIA Omniverse represents a revolutionary step forward for the industrial digital twin and physical AI in simulations,” said Nico Caprez, corporate development manager at Nvidia, in a statement . “Their ability to create scalable environments for AI training could set a new standard for industries from manufacturing to robotics.”

In 2023, MetAI collaborates with Kenmec to create digital twins for automated warehouses. MetAI technology claims to have significantly reduced the time required for warehouse digital twin simulation 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. Beyond that, the Taiwan-based startup intends to establish an office in the US and move its headquarters in the second half of 2025, Yu told TechCrunch.

“Taiwan serves as our testing ground, where we collaborate with industry leaders in Taiwan to integrate deep vertical knowledge into our models, ensuring our solutions are robust and scalable,” Liu said. . “Due to its size and the need 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 both point solutions and end-to-end solutions, including SaaS offerings and vertical AI agents designed for rapid implementation of real-world scenarios within these industries.