In his public appearances, Nvidia CEO Jensen Huang constantly drops the buzzword “Omniverse”, which is defined as an “extensible platform for virtual collaboration.” Applications for the platform include smart cities, robotics, self-driving cars, factory design, and assembly line manufacturing.

Robotics and smart manufacturing are expected to become critical application battlegrounds for Nvidia in the coming decade.

During his high-profile 10-day tour of Taiwan, apart from thanking suppliers, Huang sought to impart new imagination regarding AI smart factories and warehousing logistics, stating that the more comprehensive the ecosphere becomes, and the more businesses integrate into it, the more Nvidia’s computing power will be needed.

“Jensen believes that factories have considerable data for the application of generative AI scenarios,” says Johnny Chen, chairman of Solomon Technology.

“He feels that manufacturing rules the roost in Taiwan, so there is considerable opportunity,” an anonymous high-level Taiwanese manufacturing company executive told CommonWealth.

Taiwan Ideally Positioned as Omniverse Arrives

Nvidia recently announced plans to hire 1,000 workers in Taiwan. Reportedly, most of that manpower will be devoted to the company’s Omniverse division, with a focus on the needs for the digital transformation of Taiwanese industry and transition to “smart” factory operations.

Once implemented, the Omniverse can align with the multinational operations of Taiwan-based manufacturers to expand to manufacturing in other countries.

Taiwanese manufacturers are undoubtedly key players in the Nvidia ecosphere. Solomon, which was name checked by Huang in March, has seen its stock prices soar more than fourfold this year.

Solomon, which at the time was the licensed distributor for Universal Robot’s robotic arms, began training its own visual AI models seven years ago. Today, this has blossomed into an algorithm team numbering in excess of 100 people. And although AI automation only accounts for less than 10 percent of the company’s revenue, when visual applications platforms began being incorporated around the world, Nvidia came knocking.

Salomon has transformed from a distributor and now has three major product lines: robot application, AI defect detection, and AR and AI integration. (Photo: Chien-Tong Wang)

Solomon is able to use algorithms developed in-house to emulate the movement paths of machines. However, after using Nvidia’s platform, it discovered that efficiency had increased by over ten-fold. After modeling using the logistical assembly line developed in partnership with KENMEC, it was ready for direct implementation into the robot production line.

Nvidia also approached KENMEC directly. KENMEC had already been a vendor for Giant’s logistics center, handling automated warehousing and logistical systems integration. Recently, as Giant worked on plans for some new warehousing and logistics spaces, KENMEC proposed using AI on the Nvidia platform to simulate operational efficiency.

Jonas Ko, president of KENMEC. (Photo: Ming-Tang Huang)

Jonas Ko, president of KENMEC Mechanical Engineering, relates that such physical models as speed, heat transfer, and optical performance have been modeled on the Omniverse, saving a lot of back-and-forth on each tool platform.

“(Nvidia) wants to be the Microsoft of industry,” offers Sebastian Hou, managing director and senior investment analyst at Neuberger Berman. Boasting its own operating system platform along with the software bank accumulated over the years via the CUDA unified parallel computing platform, it is like having a brain and heart at the core of the system.

‘The Microsoft of Industry,’ only stronger

The difference that makes Nvidia even stronger than Microsoft is hardware and software integration. Moreover, it is an open ecosphere.

By deepening connections with customers and driving hardware demands through the ecosystem, even if the AI chip market sees price drops and increased competition in the future, Nvidia expects to be able to separate itself further from other AI chip suppliers and continue to generate revenue growth.

Nvidia’s current hardware and software advantage has compelled conventional automation giant Siemens to join its ecosystem.

Huang constantly sings the praises of the Omniverse’s Digital Twin. Siemens had talked about something along the same lines for over a decade, but “raised the digital twin to the next level creating the Industrial Metaverse,” vice president of Siemens Digital Industries Taiwan, barely concealing his excitement.

Aditya Ramkrishna, vice president of Siemens Digital Industries Taiwan. (Photo: Pei-Yin Hsieh)

“You can’t just put your shoulder to the grindstone on your own when it comes to AI; it really needs cooperative ecosystem partners,” adds Ramkrishna. Capable of simulating behavior in the physical realm through digital modeling with its Xcelerator platform, Siemens announced a cooperative deal with Nvidia’s Omniverse, which will help AI understand such things as weight, time, temperature, and drag coefficient.

“If an error happens on site, it can immediately be traced back using the Digital Twin to identify and correct the problem,” relates Ramkrishna.

"People see New Money, not just New Cost," observes Jerry Huang, CEO of Profet AI.

Profet AI has over 200 clients in Taiwan, and is one of nearly 20,000 members of NVIDIA's Inception startup incubation platform. The company specializes in using machine learning to integrate AI with human experience and machine data, applying it in areas such as factory expansion and personnel training.

Jerry Huang points out that the future digital transformation of manufacturing plants could use the Nvidia platform as the foundation, layering new startup products on top of it depending on demand. This also conforms with Nvidia’s strategy to strengthen its ecosystem.

These startup teams, like Jensen Huang, firmly believe that AI can accelerate the evolution of manufacturing plants.

Dr. Paul Shieh, founder and CEO of Linker Vision, founded a company in the USA, which was acquired by Cisco for US$2.9 billion 15 years ago. He returned to Taiwan in 2015 to start a business focused on AI, initially developing visual recognition machine learning for inspections and autonomous driving. Five or six years ago, when Bosch, Samsung, and LG were developing autonomous driving businesses, they approached his 200-person team in Tainan to start with basic functions such as image data tagging to train large visual models. 

Dr. Paul Shieh, founder and CEO of Linker Vision (Left). (Photo: Ming-Tang Huang)

Globally, Nvidia’s Developer Relationship Department acts as a kind of talent scout, constantly seeking promising startups. Last year, they approached Linker Vision.

“They were quite surprised to find a software company in Taipei doing such a good job developing a large visual model platform,” recalls Shieh. Nvidia discovered that AI services in enterprises can take a long time from proposal to implementation - sometimes eight or nine months just to gather data. “This is a big issue for AI advancement.”

They came to Linker Vision because the company’s visual model technology could accelerate the implementation of services on the Omniverse platform, especially applied to factories, where it can identify the attributes of moving objects in real-world images and link directly to digital twins.

Nvidia took Linker Vision into Inception, and introduced them to become a cooperative partner with KENMEC.

Even a small company without an office can join

Another very high profile startup company is MetAI, whose technology is able to rapidly translate 2D plans into 3D models.

MetAI co-founder and chief technology officer Renton Hsu was working in Hollywood as a 3D special effects technician. Realizing that AI was coming for his job, he pivoted and went into AI. While working as an engineer at an artificial intelligence school in Taiwan, he developed a 3D modeling solution aimed at the manufacturing industry, earning him top honors in an Nvidia-sponsored competition. As a result, Nvidia encouraged him to start a business.

At present, the new company is still looking for office space.

MetAI co-founder and chief technology officer Renton Hsu(Right), CEO Tai-Wan Yu(Middle), COO Da-Xiong Liu (Left). (Photo: Ming-Tang Huang)

Using MetAI's technology, AI-modeled virtual factories can undergo multiple iterations. A factory can determine how many robotic arms to buy and how many AMRs to deploy. Through iterative modeling, the effectiveness of different brand combinations can be compared, helping business owners find the best solution.

Underlying concern: Could partners become competitors?

Jensen Huang repeatedly stresses Nvidia’s desire to create “AI contract manufacturing factories.” Just as TSMC produces chips for the entire world, Nvidia can similarly serve various application scenarios.

Still, quite a few ecosphere vendors relate an underlying concern that, as the Omniverse absorbs knowledge from the professional realm, could it turn the tables on current partners and make them into competitors?

This concern is also shared by numerous Taiwanese corporations, who are approaching cooperation with Nvidia with caution.

Have you read?

• On New TSMC Chairman’s Mind: AI, Geopolitics, Global Expansion
• Intel's Gelsinger Has a Plan to Catch Up with TSMC
• TSMC’s New Chairman C.C. Wei Sports Unorthodox Leadership Style