Sources: SoftBank, Sony, Honda, and six other Japanese companies launch a new AI company to develop a ~1T-parameter foundation model for “physical AI” by 2030
TOKYO — SoftBank has established a company to develop artificial intelligence in Japan, with NEC and Honda Motor among eight peers …
Context & Ripple Effects
The venture follows SoftBank’s data-center partnership with Microsoft and Sakura Internet, linking model development to a domestic compute buildout rather than treating AI as a stand-alone software project.
It also became the nucleus for a broader industrial program: the same company group later received government backing for the Noetra consortium and was tied to a planned procurement of Nvidia Rubin chips for robotics-oriented model work.
First-order effects
- SoftBank, Sony, Honda, NEC and the other participants gain a shared vehicle for developing a large foundation model aimed at physical-world applications, pooling their AI, industrial and deployment interests.
- The effort gives the participating manufacturers a route to shape a common model layer for vehicles, robotics and other embodied systems rather than relying solely on externally supplied models.
Second-order effects
- A model program of this scale raises demand for Japan-based data-center capacity, advanced accelerators and AI engineering talent, reinforcing the infrastructure commitments already associated with SoftBank and its partners.
- Japanese automakers and industrial-technology rivals face greater pressure to decide whether to join interoperable domestic AI efforts or fund proprietary stacks; the separate Toyota–NTT automotive AI plan shows that competing industry platforms are already forming.
Third-order effects
- If the consortium converts shared research into deployable systems, Japan’s AI competition could shift from individual corporate pilots toward nationally supported, sector-specific foundation-model platforms tied to local compute and hardware supply.
- The model’s eventual value will depend less on parameter count alone than on whether member companies can contribute differentiated industrial data, safety validation and real deployment channels—capabilities that are difficult to assemble through cloud access alone.
The trend: This is part of AI industrialization: countries and incumbent manufacturers are combining compute, capital and domain-specific deployment channels to build foundation models for physical industries.