Japan's GPU Cloud Imperative: A National Security Priority
Japan is executing the most ambitious sovereign GPU infrastructure buildout in Asia. Combined government and private sector commitments now exceed ¥10 trillion ($65 billion) through 2030, with an additional $70 billion from major technology corporations according to analysis from Introl. This is not an aspiration — it is a funded, operational, multi-year deployment with GPU hardware arriving at data centers across the country's central, northern, and western regions. The strategic driver is unambiguous: Japan identified its near-total dependency on American hyperscalers for AI compute as a critical vulnerability in its economic security architecture, and it is spending accordingly to close that gap.
The urgency is quantifiable. Japan's digital trade deficit reached ¥5.5 trillion ($36 billion) in 2023 — meaning Japanese enterprises are sending more money abroad for digital services than they receive. Every AI model trained on AWS or Azure represents a capital outflow and a strategic dependency. Every GPU cloud server deployed on Japanese soil by a METI-certified domestic provider reverses that equation. This is not merely technology policy — it is balance-of-payments economics applied to AI infrastructure.
What makes Japan's approach distinctive among sovereign AI programs is the explicit integration of GPU compute sovereignty with broader economic security legislation. The Ministry of Economy, Trade and Industry (METI) is deploying subsidies through the Economic Security Promotion Act — the same legal framework governing semiconductor supply chains, critical minerals, and defense technology. GPU cloud infrastructure has been formally classified as a "specified critical product" requiring domestic supply security. In global technology policy, Japan has elevated cloud computing to the same strategic tier as semiconductors and rare earths.
Japan's urgency stems from a structural digital trade deficit of ¥5.5 trillion ($36 billion) in 2023 — capital flowing to US hyperscalers for cloud and AI services rather than building domestic capability. METI has warned of a "2025 Digital Cliff" where unmigrated legacy systems risk annual economic losses measured in trillions of yen. The government's response: the largest sovereign compute investment program in Asia, designed to make Japan self-sufficient in AI infrastructure while positioning the nation as Asia's premier AI-friendly regulatory environment.
Market Intelligence: The Scale of Japan's AI Infrastructure Investment
Japan's AI infrastructure investment operates at three distinct tiers, each with different funding mechanisms, timelines, and strategic objectives. The total commitment — government and private combined — represents one of the largest coordinated technology infrastructure investments in Asia-Pacific history.
Tier 1: Government direct funding. The Japanese government's core AI commitment is ¥2 trillion ($13.2 billion) allocated for 2024–2025, flowing through multiple channels: ¥1.05 trillion for next-generation chip and quantum computing research, ¥471.4 billion for domestic advanced chip production, and ¥72.5 billion ($470 million) awarded directly to five companies for AI supercomputer development through METI. The fiscal 2025 budget proposal includes ¥330 billion for AI and semiconductor sectors, with ¥100 billion specifically allocated to Rapidus for 2nm semiconductor production.
Tier 2: METI GPU cloud subsidies. In 2025, METI expanded its subsidy program, allocating approximately $740 million (¥114.6 billion) to six certified firms for sovereign cloud infrastructure development. Sakura Internet received the largest allocation at ¥50.1 billion ($324 million), followed by KDDI at ¥10.2 billion ($66 million). Additional subsidies went to GMO Internet Group, Highreso, Rutilea, and Ubitus.
Tier 3: Private sector deployment. NTT Corporation leads domestic private investment with $59 billion (¥8 trillion) over five years through 2027, including a $16.4 billion buyout of NTT Data to consolidate AI strategy and nearly 1GW of new global data center capacity. SoftBank Corp. announced ¥150 billion ($960 million) specifically for AI infrastructure expansion including NVIDIA GPU purchases. The Japan cloud computing market overall is expected to grow by $31.3 billion from 2026–2030 at a CAGR of 18.6% according to Technavio.
METI Subsidy Architecture: How the Money Flows
The METI subsidy program for sovereign AI cloud infrastructure operates under the Economic Security Promotion Act, which classifies cloud computing alongside semiconductors, critical minerals, batteries, and pharmaceuticals as specified critical products requiring domestic supply security. This legal framework gives the subsidies a permanence and institutional backing that discretionary budget allocations would not — they are embedded in Japan's economic security legislation, not subject to annual budget negotiation.
METI funds up to 50% of the total investment made by certified companies, with the remainder coming from private capital. The certification process requires companies to demonstrate technical capability, data center infrastructure, NVIDIA partnership credentials, and operational readiness. As of 2025, METI has certified companies under both the original 2024 program and an expanded 2025 program that increased the number of providers from five to six and nearly doubled total funding from ¥72.5 billion to ¥114.6 billion.
The subsidy structure creates a powerful incentive alignment: domestic providers receive government capital to build infrastructure that serves both commercial and national security objectives. For enterprise procurement officers evaluating GPU cloud vendors in Japan, METI certification is a critical signal — it indicates financial backing, government validation, and long-term operational commitment. The certified providers are not competing on venture capital runway — they are competing on sovereign infrastructure mandates.
METI awarded ¥72.5 billion ($470 million) to five companies for AI supercomputer development under the Economic Security Promotion Act. SAKURA Internet received the largest allocation at ¥50.1 billion ($324 million), followed by KDDI at ¥10.2 billion. This funding addresses Japan's ¥5.5 trillion digital trade deficit (2023) by keeping sovereign AI compute spending domestic rather than flowing to US hyperscalers.
The ¥72.5 billion METI allocation was certified under the Economic Security Promotion Act, designating cloud compute as "specified critical products" essential to national security. SAKURA Internet received ¥50.1 billion ($324 million), KDDI received ¥10.2 billion ($66 million), and smaller allocations went to GMO Internet Group, Highreso, and Rutilea. METI's fiscal 2025 budget proposal includes an additional ¥330 billion for AI and semiconductor sectors, with ¥100 billion specifically for Rapidus's Hokkaido semiconductor factory targeting 2027 mass production.
The Six Providers: Japan's GPU Cloud Ecosystem
Six companies have been certified by METI to build Japan's sovereign GPU cloud infrastructure, each operating in distinct geographic regions and serving different market segments. Understanding this ecosystem is essential for enterprise GPU procurement decisions in Japan.
Sakura Internet (¥50.1B subsidy). Sakura is the largest METI subsidy recipient and Japan's most established domestic cloud provider. The company is expanding from 2,000 to approximately 10,800 GPUs, installing NVIDIA HGX B200 infrastructure at its Ishikari data center in Hokkaido. Sakura's customers include the National Institute of Informatics (NII), which uses the platform for Japanese-centric medical LLM development. In April 2025, Sakura signed a basic agreement with KDDI and Highreso for mutual GPU capacity sharing to address demand surges.
SoftBank Corp. (major private investment). SoftBank operates the world's first NVIDIA DGX SuperPOD with DGX B200 systems, targeting 10,000+ GPUs delivering 13.7 exaflops of AI compute. Its subsidiary SB Intuitions uses NVIDIA AI infrastructure for Japanese-native large language model development. In October 2025, SoftBank launched Cloud PF Type A, a sovereign cloud platform powered by Oracle Alloy, providing 200+ OCI services with encryption key custody under SoftBank's proprietary Key Management Service.
KDDI (¥10.2B subsidy). KDDI, Japan's second-largest telecommunications carrier, is building liquid-cooled data centers featuring the NVIDIA GB200 NVL72 platform with Grace Blackwell Superchips. KDDI's approach integrates GPU cloud with its existing 5G and telecommunications infrastructure, offering unique advantages for low-latency AI inference at the network edge.
GMO Internet Group. GMO launched GMO GPU Cloud, the first local cloud in Japan featuring full-stack NVIDIA H200 Tensor Core GPUs with NVIDIA Spectrum-X Ethernet networking and BlueField-3 DPUs. Built on Dell PowerEdge servers with NVIDIA AI Enterprise software suite, GMO GPU Cloud targets enterprises building generative AI applications.
Highreso. Based in Kagawa Prefecture, Highreso operates GPU data centers at closed schools — a model for regional economic revitalization through AI infrastructure. The Kagawa facility, certified under the METI Cloud Program, was the first GPU-dedicated data center in the Chugoku/Shikoku regions. Highreso's GPUSOROBAN platform has been utilized over 2,000 times across IT, manufacturing, construction, and university research.
Ubitus. Taiwan-based Ubitus received a 2025 METI Growth Investment Grant to deploy next-generation Blackwell-based NeoCloud architecture across multiple Japanese regions. Ubitus focuses on AI applications for tourism, culture, education, and healthcare, with specialized Japanese language models.
In April 2025, KDDI, SAKURA Internet, and Highreso signed a basic agreement to jointly meet surging GPU demand, creating a reciprocal usage system where the three providers can flexibly allocate GPU capacity across their respective data centers. All three are certified under METI's Economic Security Promotion Act. Highreso opened Japan's first GPU-dedicated data center in the Chugoku-Shikoku region in Kagawa (December 2024) and is expanding to repurposed closed schools in Saga and Kagawa — a creative approach to regional revitalization through sovereign AI infrastructure.
NVIDIA Japan: The Strategic Infrastructure Partnership
NVIDIA CEO Jensen Huang has personally driven the Japan sovereign AI relationship through meetings with government leaders and a keynote at the NVIDIA AI Summit Japan. The partnership architecture positions NVIDIA not merely as a GPU vendor but as a co-architect of Japan's sovereign AI infrastructure. All six METI-certified providers deploy NVIDIA accelerated computing, networking, and software — making NVIDIA the common technology stack across Japan's entire domestic GPU cloud ecosystem.
Key NVIDIA platforms deployed across Japan include the DGX B200 (SoftBank SuperPOD), HGX B200 (Sakura Internet), GB200 NVL72 with Grace Blackwell Superchips (KDDI), H200 Tensor Core GPUs (GMO), and Hopper GPUs (Rutilea, Highreso). The NVIDIA AI Enterprise software suite provides a common application development layer across all providers. For enterprise architects, this means GPU cloud workloads developed on one METI-certified provider are technically portable to others — reducing vendor lock-in risk within the domestic ecosystem.
NVIDIA CEO Jensen Huang announced at the NVIDIA AI Summit Japan that six Japanese cloud leaders — SoftBank, GMO Internet Group, Highreso, KDDI, Rutilea, and SAKURA Internet — are building AI infrastructure with NVIDIA accelerated computing to transform robotics, automotive, healthcare, and telecom. SAKURA Internet plans to install NVIDIA HGX B200 Blackwell infrastructure at Ishikari, targeting 10,800 GPUs powered entirely by renewable energy by 2027. KDDI is planning a liquid-cooled data center featuring the NVIDIA GB200 NVL72 platform with Grace Blackwell Superchips at its Osaka Sakai facility.
SoftBank's Sovereign AI Buildout: The Scale Player
SoftBank's GPU infrastructure investment deserves separate analysis because of its scale, its sovereign cloud dimension, and its implications for the competitive landscape. The 10,000+ GPU DGX SuperPOD — the world's first deployment of DGX B200 at this scale — delivers a combined 25.7 exaflops of total compute capability when fully deployed. This positions SoftBank as the single largest GPU compute operator in Japan and one of the largest outside the United States.
The sovereign cloud dimension is equally significant. SoftBank's October 2025 partnership with Oracle to launch Cloud PF Type A on Oracle Alloy creates a full sovereign cloud stack — not just GPU compute but 200+ cloud services including database, analytics, AI, and enterprise applications, all hosted in SoftBank's eastern and western Japan data centers with SoftBank's own proprietary Key Management Service controlling encryption keys. This gives SoftBank a structural advantage over pure GPU compute providers — it offers an integrated sovereign platform where enterprises can run entire application stacks without data leaving Japanese jurisdiction.
SoftBank operates the world's first NVIDIA DGX SuperPOD with DGX B200 systems, targeting 10,000+ GPUs delivering 13.7 exaflops and eventually 25.7 exaflops total compute. SoftBank has committed an additional ¥2 trillion ($12.7 billion) of its own capital for data center infrastructure, and its ownership of Arm Holdings gives it vertical influence from chip architecture to foundation models — a unique positioning in sovereign AI.
SoftBank operates the world's first NVIDIA DGX SuperPOD with DGX B200 systems, targeting 10,000+ GPUs delivering 13.7 exaflops. SoftBank committed an additional ¥2 trillion ($12.7 billion) for data center infrastructure, and its Arm Holdings ownership gives it vertical influence from chip architecture to foundation models — unique positioning in sovereign AI globally.
ABCI 3.0 & National Supercomputing Infrastructure
The National Institute of Advanced Industrial Science and Technology (AIST) operates Japan's public AI supercomputing infrastructure through the AI Bridging Cloud Infrastructure (ABCI) program. ABCI 3.0, funded with ¥36 billion ($232 million) through METI's Economic Security Fund, is the third generation of this infrastructure and has been operational since January 2025.
Built by Hewlett-Packard Enterprise and powered by NVIDIA GPUs, ABCI 3.0 delivers a theoretical peak performance of approximately 6.2 exaflops (half-precision), making it the most powerful public supercomputer in Japan. AIST offers ABCI 3.0 access to companies and universities for foundational model development — functioning as shared national AI compute infrastructure that democratizes access to scale that individual companies could not afford independently.
For the investment community, ABCI 3.0 represents a public infrastructure play that complements the private GPU cloud buildout. Companies developing AI models in Japan can use ABCI 3.0 for training and research, then deploy production workloads on commercial METI-certified providers. This two-tier architecture — public research compute plus commercial production compute — mirrors successful models in semiconductor R&D and positions Japan to accelerate AI development without requiring every startup to procure its own GPU clusters.
AIST's ABCI 3.0 — funded with ¥36 billion ($232 million) through METI's Economic Security Fund — delivers 6.2 exaflops peak performance, making it Japan's most powerful public supercomputer. Built by HPE with NVIDIA GPUs, ABCI 3.0 has been available since January 2025 for foundation model development. The National Institute of Informatics uses SAKURA Internet infrastructure to develop Japanese-centric medical LLMs through its Research and Development Center for Large Language Models.
AIST's ABCI 3.0 — funded with ¥36 billion ($232 million) through METI's Economic Security Fund — delivers 6.2 exaflops peak performance, making it Japan's most powerful public supercomputer. Built by HPE with NVIDIA GPUs, it has been available since January 2025 for foundation model development. The National Institute of Informatics uses SAKURA Internet infrastructure for Japanese-centric medical LLMs.
Investment & Capital Flows: Following the Money
The capital allocation into Japan's GPU cloud infrastructure creates investment exposure across multiple vectors that institutional investors and fund managers should evaluate.
Direct GPU infrastructure: SoftBank's ¥150 billion GPU investment, NTT's ¥8 trillion five-year commitment, and METI's ¥114.6 billion subsidy program collectively represent over $75 billion in committed capital flowing into Japanese AI infrastructure between 2024 and 2030. This capital funds GPU hardware purchases (primarily from NVIDIA), data center construction, power infrastructure, and cooling systems.
Semiconductor supply chain: Rapidus has received ¥920 billion in government support for its Hokkaido factory targeting 2nm chip production by 2027. TSMC's Kumamoto fab (JASM) represents a separate $8.6 billion investment in domestic semiconductor manufacturing. These investments secure the physical chip supply chain that underpins GPU cloud infrastructure.
Public market exposure: Sakura Internet (TSE: 3778), KDDI (TSE: 9433), SoftBank Corp. (TSE: 9434), GMO Internet (TSE: 9449), and NTT (TSE: 9432) are all publicly traded on the Tokyo Stock Exchange. Each company's GPU cloud strategy represents a quantifiable percentage of revenue growth and capital expenditure that analysts can model. Sakura Internet, as the largest METI subsidy recipient, offers the most direct pure-play exposure to sovereign GPU cloud deployment.
Service layer opportunities: System integration, AI model development, managed GPU services, and enterprise consulting represent high-margin service opportunities built atop the hardware infrastructure. Japanese consulting firms and global system integrators with Japan practices stand to capture significant revenue as enterprises migrate AI workloads to domestic providers.
Semiconductor Sovereignty: The Hardware Foundation
Japan's GPU cloud strategy is explicitly linked to its semiconductor sovereignty program. The logic is straightforward: sovereign AI compute is only truly sovereign if the chips powering it can be sourced domestically or from allied supply chains immune to adversary disruption. Two flagship programs address this dependency.
Rapidus Corporation, backed by ¥920 billion in government funding, is building Japan's first leading-edge logic fab in Chitose, Hokkaido, targeting 2nm chip production by 2027. Rapidus partners with IBM for process technology and aims to manufacture chips for AI inference, automotive, and defense applications. If successful, Rapidus would give Japan domestic access to cutting-edge semiconductor production for the first time in over a decade.
TSMC's JASM (Japan Advanced Semiconductor Manufacturing) fab in Kumamoto — a $8.6 billion investment with Japanese government subsidies covering approximately 40% — began production of 12/16nm and 22/28nm chips in 2024, with plans to expand to 6nm. While these are not leading-edge AI training chips, they serve essential roles in automotive, IoT, and edge AI applications. A second Kumamoto fab is planned for more advanced nodes.
For enterprise decision-makers, the semiconductor dimension means Japan's sovereign GPU cloud strategy has a longer time horizon than just current NVIDIA GPU deployments. By 2028–2030, Japanese AI chip alternatives — including Preferred Networks' MN-Core custom AI accelerator and Rapidus-manufactured inference chips — may reduce the ecosystem's current near-total dependency on NVIDIA hardware.
Japan has committed ¥920 billion to Rapidus for its Hokkaido fab targeting 2-nanometer chips by 2027, with an additional ¥1.05 trillion ($7 billion) for quantum computing and next-generation chip research. The NTT-led consortium of NTT, KDDI, Fujitsu, NEC, and Rakuten is developing Beyond-5G PaaS integrated stacks. SoftBank's ownership of Arm Holdings gives Japan vertical influence from chip architecture to foundation models — a unique positioning in sovereign AI that no other nation replicates. Japan's immersion cooling innovations are achieving PUE of 1.02-1.04, with KDDI's container-type solution delivering 43% power reduction.
Sovereign Cloud: Oracle Alloy & Data Residency
Japan's sovereign cloud dimension extends beyond GPU compute into full-stack cloud platforms meeting Japanese data residency and regulatory requirements. SoftBank's Cloud PF Type A on Oracle Alloy and Sakura Internet's infrastructure both ensure that enterprise data — including AI training data, model weights, and inference results — remains within Japanese jurisdiction, processed by Japanese entities, under Japanese law.
Japan's Act on the Protection of Personal Information (APPI) and its 2022 amendments establish cross-border data transfer restrictions that directly impact AI workload placement. Financial institutions regulated by the Financial Services Agency (FSA) and healthcare organizations under Ministry of Health guidelines face additional data residency requirements. The practical consequence is that regulated Japanese enterprises cannot train AI models on foreign cloud infrastructure without navigating complex adequacy determination processes — making domestic sovereign GPU cloud the path of least compliance resistance.
ISMAP (Information System Security Management and Assessment Program) is Japan's government cloud security certification, operated jointly by NISC, the Digital Agency, MIC, and METI. AWS, Microsoft Azure, Google Cloud, and IBM Cloud are all ISMAP-registered, alongside domestic providers. Japanese government agencies must procure cloud services from ISMAP-registered providers — creating a structured certification market analogous to FedRAMP in the US or SecNumCloud in France. ISMAP-LIU (Low-Impact Use) extends the framework to SaaS services, with Japan's Digital Agency actively promoting registration.
Risk Assessment & Strategic Vulnerabilities
Despite the scale of investment, Japan's sovereign GPU cloud strategy faces identifiable risks that investors and procurement officers should evaluate.
IT labor shortage. METI projects Japan will face a shortfall of 3.2 million workers in AI and robotics technologies by 2040. Low salary levels relative to Silicon Valley drive brain drain, and Japanese universities produce fewer AI/ML doctoral graduates than competitors. This talent constraint may limit the rate at which deployed GPU capacity can be utilized effectively.
Training data scarcity. Japanese language text data for AI model training is substantially less abundant than English or Chinese corpora. The absence of major domestic internet platforms (equivalent to Facebook, X, or WeChat) limits the volume and diversity of available training data. This means sovereign GPU compute infrastructure may be underutilized unless training data acquisition strategies improve.
NVIDIA dependency. All six METI-certified providers deploy NVIDIA hardware exclusively. While domestic alternatives (Preferred Networks MN-Core, Rapidus inference chips) are in development, they will not reach commercial scale until 2028 at earliest. Any disruption to NVIDIA supply — whether from export controls, production constraints, or geopolitical events — would impact the entire ecosystem simultaneously.
Power infrastructure. GPU-dense data centers require substantially more power than traditional facilities. Japan's Energy Saving Act mandates PUE of 1.4 or lower, and leading facilities achieve PUE 1.2–1.3 with liquid immersion cooling reaching PUE 1.02–1.04. However, Japan's electricity costs remain higher than competitors like the UAE ($0.05–0.06/kWh) and the United States, creating ongoing operational cost pressure.
Japan faces a projected IT worker shortfall of 3.2 million by 2040 according to METI — a structural constraint that limits sovereign AI development regardless of capital investment. Japanese-language training data is scarce compared to English or Chinese, and the absence of major domestic internet platforms (no Japanese equivalent of Facebook or WeChat) limits the data available for foundation model training. The sovereign AI initiative addresses these constraints through simplified visa tracks for foreign AI talent and massive university research funding, but closing the talent gap with the US and China remains Japan's greatest challenge.
Decision Framework: Enterprise GPU Procurement in Japan
For enterprises evaluating GPU cloud procurement in Japan, the decision matrix centers on several key variables: workload classification (training vs. inference), data residency requirements, latency sensitivity, budget constraints, and regulatory compliance profile.
METI-certified providers should be the first evaluation tier for any organization with data residency requirements, government contracts, or sensitive IP. Sakura Internet offers the largest subsidized capacity and national research institution partnerships. SoftBank offers the most complete sovereign stack (GPU + Oracle Alloy cloud services) with the highest raw compute scale. KDDI offers unique telecommunications integration for edge AI use cases.
Cost optimization: METI subsidies effectively reduce provider infrastructure costs by up to 50%, and these savings flow partially to enterprise pricing. Domestic GPU cloud pricing in Japan is trending competitive with hyperscaler equivalents for sovereignty-compliant workloads — a significant shift from even 18 months ago.
Hybrid architecture: Most enterprises will adopt a tiered model: sovereign METI-certified GPU cloud for regulated data, training data, and production AI inference; hyperscaler (AWS, Azure, Google Cloud) for non-sensitive workloads requiring global scale. The key architectural decision is where the sovereignty boundary falls in the data pipeline.
The hyperscale data center market in Japan reached $5.35 billion in 2025 and is forecast to reach $11.50 billion by 2031 at 13.58% CAGR. AWS leads capital intensity with a ¥2.26 trillion pledge through 2027. Oracle's $8 billion sovereign cloud build targets regulated clients through alliances with Fujitsu and NRI. The top five cloud vendors account for roughly 58% of total Japanese cloud spend, indicating moderate concentration and genuine opportunity for domestic providers. AI/ML workloads are growing at 22.6% CAGR through 2031, outpacing every other workload category.
Strategic Outlook 2026–2030: Japan as Asia's AI Infrastructure Hub
Japan's GPU cloud buildout is entering the operational deployment phase where installed capacity begins generating commercial revenue. By 2027, the combined GPU fleet across METI-certified providers will exceed 50,000 advanced NVIDIA GPUs — positioning Japan as the largest sovereign AI compute market in Asia outside China. The total addressable market for GPU cloud services in Japan is projected to reach $15–20 billion by 2030, driven by enterprise AI adoption, government digitalization, and the automotive industry's shift to AI-driven autonomous systems.
Several convergent trends will shape this market through 2030. First, Japanese enterprise AI adoption is accelerating as domestic GPU cloud removes the compliance barriers that slowed hyperscaler adoption. Second, Japan's positioning as an "AI-friendly" regulatory environment — in deliberate contrast to the EU's AI Act compliance burden — will attract foreign AI companies seeking a sovereign-compliant but innovation-permissive operating environment. Third, the semiconductor sovereignty investments (Rapidus, TSMC Kumamoto) will begin yielding results, gradually reducing the ecosystem's dependency on imported chips and creating a more vertically integrated Japanese AI stack.
For institutional investors, the Japan GPU cloud thesis is uniquely compelling: government-funded infrastructure buildout ($740M+ in METI subsidies), private sector capital at unprecedented scale (¥8T from NTT alone), NVIDIA strategic partnership with personal CEO engagement, and structural demand drivers (data residency regulation, digital trade deficit reduction, economic security legislation) that ensure long-term utilization. The infrastructure is being built. The demand is legislated. The question for market participants is not whether to engage with Japan's sovereign GPU cloud — it is how quickly and through which vectors.
Japan's National AI Basic Plan, formalized late 2025, commits ¥1 trillion ($6.34 billion) over five years — the nation's largest sovereign technology investment. The plan targets trillion-parameter Japanese-language foundation models through a consortium led by SoftBank. Japan faces a projected 3.2 million IT worker shortfall by 2040, addressed through simplified visa tracks for AI talent and massive university research funding.