SoftBank Bets on Physical AI With National-Scale Effort, Market Fears Fragmented Demand
AI Podcast
Approximately 30 Japanese companies, including SoftBank, NEC, Honda, and Sony, are forming a joint venture to develop "Physical AI" by 2027. This initiative aims to integrate real-world manufacturing data into trillion-parameter models, differentiating from large language models. SoftBank is investing one trillion yen in a GPU data center to support this. The venture serves as a hedge for SoftBank’s OpenAI investments and leverages Japanese government subsidies. Market concerns exist regarding fragmented company needs and data integration challenges. Investors should monitor METI's third-quarter 2026 budget approval for subsidy implementation.
TradingKey - According to Japanese media reports on May 28, approximately 30 Japanese companies are considering participating in a SoftBank-led AI joint venture, "Japan AI Foundation Model Development Co.," which aims to develop "Physical AI."
SoftBank, NEC, Honda, and Sony each hold stakes of over 10%. The first batch of about 10 companies is expected to complete their capital contributions in June, with each investing tens of millions of yen. The joint venture aims to develop a trillion-parameter model by 2027 and integrate real-world data such as temperature and weight by the early 2030s to create "Physical AI" capable of sensing and controlling the physical world.
Unlike large language models that output text, "Physical AI" relies on real-world manufacturing data. Japan's massive accumulation of process data in fields such as precision machining and automated production lines serves as a unique advantage. Furthermore, SoftBank is building a large-scale AI data center at the site of Sharp's former Sakai plant, planning to deploy approximately 100,000 GPUs with a total investment of about 1 trillion yen. Expected to be fully operational by 2028, this facility will provide sufficient computing power for training Physical AI models.
From "Lone Wolf" to "Alliance": The Logic Behind SoftBank’s In-House Development
Currently, Japan lags far behind China and the U.S. in the field of large models, and SoftBank has previously participated indirectly, primarily through investments in OpenAI. Its current pivot toward in-house development is driven by two core reasons: first, to create a "fallback plan" for its investment in OpenAI , According to media reports, SoftBank holds approximately a 13% stake in OpenAI, with cumulative investments exceeding $60 billion. However, OpenAI's valuation has surpassed $850 billion, and it faces competition and governance risks. Positioning itself in physical AI acts as a hedge for its massive capital; if OpenAI's growth slows, SoftBank has another path. Second, Japanese government subsidies reduce the cost of trial and error. Japan’s Ministry of Economy, Trade and Industry (METI) plans to provide subsidies of up to 1 trillion yen, which happens to offer a low-cost implementation opportunity for SoftBank’s "alternative plan."
Masayoshi Son, who has long been known for his "lone wolf" style of making heavy bets, has partnered with 30 Japanese companies this time. The reason is that physical AI relies on manufacturing process data, which is scattered across countless factories. Companies are unwilling to sell their core data and prefer equity swaps instead. Only through equity ties will these companies open up their production line data. At the same time, the joint venture structure is also a necessary condition for applying for government subsidies.
Market concerns over demand fragmentation.
SoftBank shares closed down approximately 2% on the day of the joint venture announcement. Market concerns were primarily centered on the fragmented requirements of various companies (Honda seeks autonomous driving, while Sony focuses on robotics), making it difficult for a single foundation model to cater to all; furthermore, while Japanese manufacturing data is abundant, standards are inconsistent, posing significant integration challenges.
On the surface, SoftBank's 1 trillion yen investment in data centers is staggering, but if Physical AI eventually becomes the industrial standard for the Japanese manufacturing sector, SoftBank could realize returns many times its initial outlay. This represents a bet with "limited downside and significant upside," rather than an all-or-nothing gamble. Therefore, investors should focus on the budget approval results from Japan's Ministry of Economy, Trade and Industry (METI) in the third quarter of 2026, which will directly determine whether subsidies can be implemented in a timely manner.
This content was translated using AI and reviewed for clarity. It is for informational purposes only.
Recommended Articles
Comments (0)
Click the $ button, enter the symbol, and select to link a stock, ETF, or other ticker.