Japan aims to build the "Siasun Robot & Automation national team", planning to purchase 27,500 NVIDIA Rubin! AI chips. The super cycle of AI computing power is moving from the cloud to "physical AI".

The Japanese government and a consortium of major Japanese corporations have jointly established a new company called Noetra Corp to purchase 27,500 next-generation AI GPUs from NVIDIA to build a massive AI computing infrastructure cluster using the cutting-edge AI computing power product NVIDIA Rubin architecture AI GPUs. This will accelerate the development of domestic Siasun Robot & Automation basic artificial intelligence models in Japan and support the creation of a large-scale Siasun Robot & Automation cluster led by the Japanese government with the support of policies and top companies. Noetra Corp, the newly established company, is essentially a "national team" project funded by the Japanese government and jointly established and operated by major Japanese corporations. Noetra Corp, established by SoftBank, Preferred Networks, NEC, Fujitsu, and dozens of other companies, is responsible for building large data centers and integrating the AI research capabilities of various companies. They plan to purchase 27,500 next-generation AI GPUs using the Rubin architecture from NVIDIA to develop domestic basic models in Japan and ultimately create a "sovereign physical AI" system for Siasun Robot & Automation; the Japanese government mainly plays the role of providing financial support and driving industrial policies. The purchase of 27,500 Rubin accelerators ignites the "Japanese Siasun Robot & Automation national team." Together with NVIDIA CEO Huang Renxun's collaboration with old industrial giants such as Fanuc, Yaskawa Electric, Kawasaki Heavy Industries, and Fujitsu in Japan to promote physical AI cooperation, and the inclusion of SoftBank, NEC, Hitachi, Sony, Preferred Networks in the Cosmos ecosystem, as well as Taiwan Semiconductor Manufacturing Company, the exclusive manufacturer of NVIDIA AI chips - which has been dubbed the "king of chip manufacturing" - recently announced strong performance exceeding expectations and an increasingly optimistic outlook for the future of AI computing power demands. This proves that the global AI computing investment cycle is far from over. The expansion of computing power is transitioning from a single arms race dominated by American super-scale cloud providers to a multi-center resonance of "cloud AI, sovereign AI, and physical AI." The bold purchase of 27,500 Rubins! The government invests 387.3 billion yen in physical AI, while Japanese giants, along with NVIDIA, are working together to build the foundation for Siasun Robot & Automation AI national base. Japanese corporate giants are planning to purchase 27,500 next-generation Rubin AI GPUs and invest at least around $2.4 billion to build domestic basic models and large data centers. This means that GPU demand is no longer solely dependent on commercial clients like Microsoft and Meta, but also receives financial support driven by national security, technological autonomy, and demographic pressures. Rubin is not just an isolated AI chip product from NVIDIA; it is an AI factory platform that integrates GPU, CPU, NVLink, networking, and DPU at the rack level. Japan's purchase will also drive demand for comprehensive AI data center infrastructure, including HBM/DRAM/NAND storage chips, advanced packaging, high-speed optical interconnects/optical communications, data center power equipment, and liquid cooling systems. The newly established Noetra company will coordinate this project and build a large artificial intelligence data center. By March next year, the new company established by Japanese giants will receive substantial government support amounting to 387.3 billion yen (approximately $2.4 billion) in government funds. Dozens of companies including SoftBank, supported by legendary investor Masayoshi Son, and Preferred Networks supported by Toyota Motor Corporation, and NEC are assisting in the formation and efficient operation of Noetra. Japan's latest AI computing infrastructure order is substantial, but compared to American tech giant Microsoft's plans to build a massive data center equipped with tens of thousands of NVIDIA Vera Rubin architecture CPUs+AIs, it is still relatively small. This initiative is one of several important measures taken by the Japanese government to reduce its dependence on foreign technology and enhance national security comprehensively. Noetra's President Makoto Danbo stated that Japan has some of the world's largest industrial-scale Siasun Robot & Automation manufacturers and is therefore poised to create an alternative to the AI super-systems of the United States and China. Makoto Danbo previously led the development of large language models for SoftBank's domestic market. "Our goal is to create a truly third choice - a solution that Japan and other countries can choose," Danbo said in an interview. He emphasized that the establishment of Noetra aims to integrate the scattered large AI projects of dozens of companies. The joint venture plans to launch an advanced AI model by March next year and have regular updates and iterations thereafter. Danbo said the long-term goal is to launch a super-physical AI model specifically designed for cutting-edge Siasun Robot & Automation applications. Noetra will mobilize engineers from SoftBank, Preferred Networks, NEC, and Fujitsu, all of which have developed their own exclusive AI models. SoftBank has the Sarashina large language model, Preferred Networks has PLaMo, and NEC's flagship model is called cotomi. Developing domestic physical AI models is at the core of Japan's strategy to build a leading global artificial intelligence and cutting-edge humanoid Siasun Robot & Automation industry center. The Japanese government's goal is to secure at least a 30% share of the global Siasun Robot & Automation market expected to reach 60 trillion yen by 2040... The global race to develop highly intelligent and advanced humanoid Siasun Robot & Automation motion systems intensifies. For Japan, in a context of declining population and severe labor shortages, driving this technology has become a very urgent task. "Japan has a lot of great ideas, but lacks labor," NVIDIA CEO Huang Renxun told reporters during an interview in Japan on Wednesday. "With automation, AI, and Siasun Robot & Automation technology, the Japanese economy is expected to thrive again." From the arms race of large models to the sovereign competition of Siasun Robot & Automation, AI investments are transitioning from the cloud to the physical world. The strategic value of Huang Renxun's visit to Japan this time is far beyond a regular client visit. NVIDIA's collaboration with Fanuc, Yaskawa Electric, Kawasaki Heavy Industries, and Fujitsu to advance physical AI cooperation, and the inclusion of SoftBank, NEC, Hitachi, Sony, Preferred Networks in the Cosmos ecosystem imply that Japan is connecting its years of accumulated Siasun Robot & Automation essence, precision manufacturing, sensors, and industrial data with NVIDIA's world models, digital twins, training platforms, and edge computing stack. The demand for AI has expanded from large language model training to Siasun Robot & Automation simulation, reinforcement learning, visual reasoning, and real-time control. Therefore, a single Siasun Robot & Automation requires both data center training computing power and edge inference computing resources, opening up a broader and longer-lasting semiconductor demand curve than traditional chat-based Siasun Robot & Automation. The reunion of Huang Renxun with his "savior" 30 years ago holds brand and historical narrative value, but what truly impacts asset pricing is NVIDIA's systematic binding of Japanese Siasun Robot & Automation essence, automotive, chip equipment, storage, materials, and optical communication supply chains. According to Sega, Huang Renxun attended an event hosted by Sega on July 15, where he appeared on stage again with former Sega president Hayao Nakayama after many years. Huang Renxun expressed his gratitude at the event, saying, "If it weren't for everything Sega did, if it weren't for everything Hayao Nakayama did, NVIDIA wouldn't have survived till today." This connection dates back to around 1996. At that time, the newly established NVIDIA faced bankruptcy after a failed project in developing graphics chips for Sega's next-generation console due to a technological route mistake. Huang Renxun admitted the failure to the then vice president of Sega, Hayao Nakayama, who chose not to blame him but instead invested around $5 million in the struggling startup. Physical AI is not about simply integrating chat models into Siasun Robot & Automation, but requires ongoing digital twin simulation, synthetic data generation, training of visual-language-motion models, reinforcement learning, and real-time inference on the edge; this will create demand for both data center training computing power and Siasun Robot & Automation edge inference. Japan has a comparative advantage in industrial Siasun Robot & Automation, precision manufacturing, and real factory data, so its AI investment logic is not to replicate the general large models of the United States, but to establish a third pole ecosystem connecting "sovereign models - Siasun Robot & Automation basic models - industrial equipment." Whether Vera Rubin will be mass-produced on schedule is a key variable for this industrial chain to move from narrative to revenue confirmation. NVIDIA has clearly stated that Rubin is entering a full-scale production phase; technically, it is not just a single GPU upgrade, but a rack-level AI factory designed to combine Vera CPU, Rubin GPU, NVLink 6, ConnectX, BlueField, and Ethernet switch chips, with the goal of reducing inference costs per token by up to 90% compared to Blackwell and completing part of MoE model training with about one-quarter of the GPUs. Huang Renxun recently denied media reports of a delay in Vera Rubin, which helps alleviate the significant tail risks of customers postponing data center, power, and network deployments, but does not fully eliminate the risks of climbing such as HBM4, complex circuit boards, liquid cooling, and rack integration; investors should not focus on whether the "chips are being produced," but rather on whether the entire rack will be delivered, customer acceptance, and stable utilization rates can be synchronized. The hype of AI investments is cooling down, but the AI computing infrastructure construction super cycle is far from over. The recent collective downturn of AI computing stocks is more like a financial pricing contraction caused by overvaluation, crowded positions, interest rate disruptions, and anxiety over investment returns, rather than a reversal of industrial capital spending. The capital market is ending the indiscriminate valuation expansion of AI assets and is moving towards the second stage of validating orders, utilization rates, and capital returns. Bank of America recently released a report showing that by 2027, under the strong trend of increasing AI inference computing power during the AI intelligent agent wave, global capital spending on cloud computing and artificial intelligence-related infrastructure is expected to reach $1.5 trillion. The current summer pullback of AI semiconductor stocks, including memory chips, is seen as a healthy reset trajectory rather than a structural change in AI computing power demand. A senior analyst team led by Brian Nowak at Morgan Stanley released a new research report on July 12, projecting significant upward revisions of capital spending forecasts for the world's five largest hyperscale cloud computing vendors (Meta, Amazon, Microsoft, Google, SpaceX) for 2027/2028 to approximately $1.2 trillion and $1.4 trillion respectively. The institution's capital spending forecast for major US tech giants in 2026 was significantly raised from $433 billion a year ago to $805 billion. Morgan Stanley's latest research revises Meta's 2027 and 2028 capital spending forecasts up by 29% and 22%, to $225 billion and $250 billion respectively; and Amazon's forecasts up by 15% and 29%, to $308 billion and $318 billion. The institution stated that the super cycle of capital spending is not over, but 2026 and 2027 may be the years with the steepest growth rates, and after 2028, stock prices will be determined not just by "who spends the most money," but by "who can most quickly convert AI computing resources into revenue, profit, and free cash flow." This new research from Morgan Stanley further emphasizes that the downturn in AI computing themes is more about deleveraging crowded positions, interest rate and geopolitical risks, and the market's reevaluation of debt financing and investment return rates, rather than any structural changes in AI computing power demand.