Son Masayoshi's 'Physical AI' Vision Nears Completion as SKAI Intelligence Breaks the Data Bottleneck

[MoneyToday]

SKAI Intelligence, a Physical AI company, has joined forces with global robotics and semiconductor leaders including ABB Robotics and NVIDIA to tackle one of the biggest bottlenecks in AI-powered manufacturing automation: shop-floor data.

According to SKAI Intelligence on July 27, the company participated as a principal contributor—alongside Deloitte and AsiaInfo—in the white paper Industrial Physical AI in High-Precision Manufacturing, jointly published by ABB Robotics and NVIDIA. SKAI Intelligence contributed its core technologies, including industrial digital twins, task-oriented synthetic data, robot vision datasets, and the Real2Sim2Real engineering loop.

The company's participation has attracted considerable attention due to ABB Robotics' strategic market position. In October last year, SoftBank Group signed an agreement to acquire ABB's robotics business for US$5.375 billion (approximately KRW 7.81 trillion). The transaction, however, remains subject to regulatory approvals in the European Union, China, the United States, and other jurisdictions, with closing expected in the second half of 2026.

The acquisition represents another major step in Masayoshi Son's long-term AI investment strategy. SoftBank has invested more than US$60 billion in OpenAI and is leading the US$500 billion Stargate AI data center project. The acquisition of ABB Robotics is widely viewed as a strategic move to secure the "physical body" that will allow AI to operate in the real world, extending beyond AI models and computing infrastructure.

In other words, following its AI ecosystem built around OpenAI (AI models), Arm (semiconductor design), and Stargate (computing infrastructure), SoftBank aims to complete the final component—a physical embodiment that enables AI to interact directly with the real world through ABB Robotics. At the SoftBank World 2026 annual conference held in Tokyo on July 14, Masayoshi Son projected that AI-related industries could generate US$46 trillion in annual revenue by 2040—equivalent to 20% of global GDP—and emphasized that achieving this vision would require approximately US$5 trillion in annual investment.

However, acquiring a robotic "body" alone does not enable robots to work like humans on factory floors. Unlike generative AI, which has advanced by learning from massive amounts of internet-based text and image data, industrial robots cannot learn manufacturing tasks solely from online data. Critical real-world variables—including lighting changes, dimensional tolerances, material properties, surface reflections, collisions, and countless edge cases—are difficult to capture at scale in actual factories, while intentionally reproducing production failures simply to collect training data is impractical.

This is why many industry experts argue that the true bottleneck in the Physical AI era is neither AI models nor robot hardware, but rather the availability of real-world experiential data required for machines to learn.

The white paper addresses precisely this challenge. Its central message is that Industrial Physical AI cannot be achieved through AI models alone. Safe deployment in manufacturing environments requires connecting digital twins, synthetic data generation, AI validation, robotic motion verification, pilot cells, and real factory feedback into a continuous engineering loop. Ultimately, success depends not on model accuracy alone, but on deployment readiness.

This is where SKAI Intelligence plays a pivotal role. In the white paper, the company is recognized as a key contributor providing industrial digital twin technology and task-oriented synthetic data. By accurately replicating manufacturing environments—including material properties, lighting conditions, camera characteristics, and industrial tolerances—in digital space, SKAI Intelligence generates high-quality synthetic datasets and automatically labeled training data for robot vision AI.

The company has also established a Real2Sim2Real engineering loop that integrates AI validation, robotic motion verification, and feedback from real production environments. This engineering framework not only improves AI model performance but also maximizes deployment readiness, enabling AI systems to be safely and reliably introduced into real manufacturing operations.

The collaboration builds on an existing strategic relationship. In June, SKAI Intelligence signed a strategic partnership agreement with ABB Robotics and has since been conducting field validation projects by integrating its ultra-high-fidelity synthetic data technology into RobotStudio, ABB's robotics simulation platform.

SKAI Intelligence is also an official NVIDIA Independent Software Vendor (ISV) partner. Earlier this year, the company completed an extended Series A financing round that valued the company at KRW 100 billion (approximately US$73 million). In addition, it signed a memorandum of understanding with the AI Institute of Seoul National University to jointly develop core robotics technologies.

A company spokesperson said, "Minister of Science and ICT Bae Kyung-hoon recently noted that advancing today's Physical AI to the level of large language models could take as long as 100,000 years if we relied solely on real-world data. This illustrates that the future competitive landscape will be determined not by AI models themselves, but by how quickly companies can generate, validate, and utilize high-quality data."

The spokesperson added, "Digital twins and synthetic data transform real-world experience into scalable data assets, dramatically shortening the time required to develop and deploy Physical AI. They will serve as critical infrastructure for the next generation of intelligent manufacturing."

Founded in November 2023, SKAI Intelligence is a Physical AI content solution company and an affiliate of SKAI, a KOSDAQ-listed technology company.