A May 8 Nikkei Scoop Reveals the Next Phase of AI Infrastructure Competition

On May 8, 2026, Nikkei broke the story: Japan’s SoftBank Corp. is in talks with NVIDIA and Foxconn to design and assemble high-performance AI server components by the end of the decade — with the goal of building “Made-in-Japan” AI compute infrastructure.

The news sent SoftBank shares down roughly 5% on the day, but that short-term jitter only underscores a more important long-term signal: Sovereign AI has moved from rhetoric into actual supply chain restructuring.

What SoftBank Is Really Doing: Three Key Intelligence Points

1. Not Just Assembly — Full In-House Manufacturing Eventually

According to follow-up coverage from Reuters citing Nikkei, SoftBank’s plan is to begin by assembling externally sourced components and then progressively take charge of the entire server manufacturing process before 2030. This is not a one-off procurement deal — it’s a ten-year industrial positioning play.

NVIDIA brings its GPU-anchored server certification program; Foxconn brings world-class AI server contract manufacturing capabilities. According to The Asia Business Daily, SoftBank is also considering the former Sharp Sakai plant — which it acquired last year — as the production base. The division of labor is clear: NVIDIA provides specs and silicon, Foxconn provides manufacturing know-how, SoftBank provides market access, land, and capital.

2. This Extends SoftBank’s “All-In Bet” on OpenAI

SoftBank Group has already invested over $30 billion in OpenAI for an approximately 11% stake — what Masayoshi Son calls an “all-in” bet. But equity alone isn’t enough. When you’re deeply tied to a frontier AI company, you also need control over the physical layer that runs those models, or you risk being squeezed by compute shortages, geopolitical disruptions, or pricing power shifts.

Building AI servers in-house is essentially extending the SoftBank–OpenAI alliance from the application layer down to the infrastructure layer.

3. “Sovereign AI” Is the Real Keyword Here

As AI adoption accelerates in administration, finance, and other sensitive sectors, demand for “Sovereign AI” — AI developed and operated domestically to prevent data leakage abroad — is rising. According to U.S. market research firm ABI Research, the AI server market is projected to reach $524 billion by 2030, nearly double its 2025 size.

What is Sovereign AI? NVIDIA’s official definition: a nation’s capabilities to produce artificial intelligence using its own infrastructure, data, workforce, and business networks. NVIDIA founder and CEO Jensen Huang has put it more bluntly: “Nobody needs atomic bombs, everyone needs AI.” In today’s geopolitical context, the statement lands harder than it sounds.

Why Now? Three Structural Forces Converging

SoftBank’s timing isn’t coincidental. Several structural pressures are converging:

Geopolitical supply chain pressure. AI servers are currently concentrated among a handful of U.S. and Taiwan-based manufacturers. For Japan — one of the most AI-strategic G7 nations — long-term reliance on foreign assembly is an obvious vulnerability.

Sovereign AI policy momentum. The Japanese government has been actively subsidizing GPU cloud infrastructure (including procurement awards to SAKURA internet and others), creating fertile policy ground for private-sector hardware investments.

OpenAI’s own hardware ambitions. Throughout 2025, OpenAI signed major GPU and custom silicon deals with AMD and Broadcom — clear signs it’s diversifying away from pure NVIDIA reliance and building its own compute stack. As OpenAI’s largest external shareholder, SoftBank has every reason to position itself at the hardware layer in parallel.

Three Things to Watch — From a Global Perspective

This isn’t just a Japan story. Three implications matter globally:

  1. “Made-in-X” may become a new compliance label for AI hardware. Just as semiconductor origin already shapes government procurement, AI servers are likely to face country-of-origin clauses in finance, healthcare, and public sector RFPs across multiple jurisdictions.
  2. Taiwan’s supply chain role evolves from component supplier to “sovereign AI manufacturing partner.” Foxconn’s deepening involvement signals that the Taiwan AI hardware industry will export full-stack AI server solutions, not just GPU boards.
  3. Sovereign AI is cascading from national to enterprise scale. When “data must stay within the country” becomes national policy, large enterprises will face the analogous internal requirement: “data must stay within the company, models must stay within the data center.” This is the single largest structural tailwind for enterprise AI infrastructure over the next five years.

SoftBank is betting on sovereign AI at the national scale. But the same logic is being reframed — at enterprise scale — across every industry where data has strategic value.

FAQ

Q1. What is “Sovereign AI”?

Sovereign AI refers to a nation’s capability to produce artificial intelligence using its own infrastructure, data, workforce, and business networks. The core principle is independence from foreign clouds and foreign models — ensuring that data, culture, and compute capacity remain under domestic control. The concept was systematically articulated by NVIDIA founder Jensen Huang at the 2024 World Governments Summit and has since become a core AI strategy pillar for Japan, Korea, India, France, the UAE, and others.

Q2. Why is SoftBank building its own AI servers?

Three main motivations: First, SoftBank has invested over $30 billion in OpenAI for an ~11% stake — building AI servers extends that bet from the application layer to the infrastructure layer. Second, Japan’s government is actively promoting sovereign AI policies, and domestic servers can capture compliance-driven procurement from government and enterprise customers. Third, the AI server market is projected to reach $524 billion by 2030, and SoftBank wants exposure to that long-term growth.

Q3. What is Foxconn’s role in this partnership?

Foxconn brings world-class AI server contract manufacturing and full-system integration capabilities. In SoftBank’s plan, Foxconn is expected to provide manufacturing know-how and supply chain orchestration, likely paired with the former Sharp Sakai plant (acquired by SoftBank last year) as the production base. This signals that Taiwan’s AI server industry is upgrading from “component supplier” to “sovereign AI hardware export partner.”

Q4. Is sovereign AI the same as enterprise on-premise AI?

Same logic, different scale. National-level sovereign AI emphasizes “data must not leave the country.” Enterprise-level on-premise AI emphasizes “data must not leave the company, models must not leave the data center.” When sovereign AI becomes national policy, large enterprises in finance, healthcare, manufacturing, and the public sector will face equivalent internal and regulatory requirements.

Q5. How big will the AI server market be by 2030?

According to ABI Research, the AI server market is projected to reach $524 billion by 2030 — nearly double its 2025 size. Three forces drive this growth: generative AI training demand, the cascade of enterprise-grade inference compute into local deployments, and national sovereign AI policies driving localized hardware procurement.

Q6. What does this mean for enterprise decision-makers worldwide?

In the short term, supply chain partners (especially Taiwan-based manufacturers like Foxconn) will deepen their role in sovereign AI hardware. In the medium term, large enterprises — especially in finance, healthcare, manufacturing, and government-adjacent sectors — will likely face board, compliance, or customer questions within three to five years: “How much of our AI compute do we actually control?” This will drive structural growth in enterprise AI compute platforms and on-premise deployment.

This article was prepared by the INFINITIX editorial team. INFINITIX’s AI-Stack platform focuses on enterprise-grade AI compute resource management and on-premise AI deployment.