SoftBank’s Industrial AI Alliance Draws 30 Japanese Giants as Japan Plans $6.7 Billion Subsidy Push

SoftBank's Physical AI consortium draws 30+ Japanese giants including Fanuc, Hitachi; targets 1-trillion parameter model by 2027 with ¥1 trillion government backing.

SoftBank’s bid to build a homegrown artificial intelligence platform for robotics is attracting corporate heavyweights at a pace that has surprised even the venture’s architects. More than 30 Japanese companies are now evaluating participation in the “Japan AI Foundation Model Development” project, a consortium that formalized in April 2026 with SoftBank Corp., NEC, Sony and Honda as founding members. The swelling ranks include industrial stalwarts such as Fanuc, Yaskawa Electric, Daikin, Asahi Kasei, Hitachi, Toshiba and Fujitsu, as well as telecoms operator KDDI.

The venture’s core focus is what the partners call “Physical AI” — systems designed to control and optimize industrial robots, factory lines and logistics in real time. Unlike general-purpose chatbots, this model will be trained on proprietary data drawn directly from Japan’s manufacturing plants, leveraging the country’s deep expertise in precision machinery and automation. The ambition is steep: the alliance aims to deliver a 1-trillion-parameter AI model by 2027, placing it alongside the most powerful global systems. A second phase slated for 2029 will add simultaneous image and audio processing, while the 2030s should see the integration of sensor data such as weight, temperature and distance.

Behind the project lies a broader national push for AI sovereignty. Japan’s Ministry of Economy, Trade and Industry (METI) is considering subsidies of up to ¥1 trillion (roughly $6.7 billion) by 2030 to accelerate domestic development, while the state-backed NEDO agency is reviewing a funding application that could shift the consortium into a public-private structure as early as this summer. For SoftBank, a government endorsement would provide both capital and regulatory cover, cementing its position as the linchpin of Japan’s industrial AI ecosystem.

Despite the strategic momentum, SoftBank’s shares have yet to reflect the optimism. The stock slipped about 2% on May 28 to close at ¥7,125, dragged lower by broader Nikkei 225 weakness tied to rising energy prices and geopolitical tensions in the Middle East. At ¥40.6 trillion in market capitalisation, the company trades on a price-to-earnings ratio of roughly 8.16 — a startlingly low multiple for a technology conglomerate with ambitions that span chips, robotics and generative AI. The discount suggests investors are treating the Physical AI play as speculative, at least until the consortium’s capital commitments and NEDO’s decision crystallise.

Meanwhile, SoftBank is reshaping its top management and preparing to unlock value from its sprawling portfolio. Rene Haas, already chief executive of chip designer Arm, has been handed an additional role as head of SoftBank Group International, tasked with coordinating the group’s semiconductor and AI holdings. The group has also mandated banks to explore US initial public offerings for its renewable energy subsidiary SB Energy and for Roze, a robotics spin?off — moves that would monetise parts of its diverse tech stable and provide fresh ammunition for its AI bets.

The alliance’s immediate test comes in June, when the new investors must formally sign on. Their likely contributions are smaller than the core members’ stakes of more than 10% each, but their involvement would embed the project deeper into Japan’s real economy. Whether SoftBank’s lowly earnings multiple eventually catches up with its technological vision may depend on how quickly those factories begin feeding data into Japan’s most ambitious AI prototype.