Not Hitachi. Not Fujitsu. Not NTT.

When Anthropic picked its first Japanese "global partner" to deploy Claude across corporate Japan, it chose NEC, a 125-year-old electronics giant better known for facial recognition, submarine cables, and government IT than for Silicon Valley-style AI plays. On April 23, 2026, the two companies announced a strategic partnership that will embed Claude Opus 4.7 and the new Claude Cowork desktop agent into NEC's DX platform, deploy Claude to roughly 30,000 employees, and co-develop AI solutions for Japanese banks, manufacturers, and local governments. The twist: NEC already has its own large language model (cotomi) and is a co-investor in Japan's national "Physical AI" foundation model company. So why add Anthropic on top, and why Claude instead of ChatGPT?

What the Deal Actually Covers

This isn't a simple reseller agreement. The NEC–Anthropic partnership has three distinct pillars, and each tells you something different about where enterprise AI in Japan is headed.

Pillar 1: Industry-specific AI solutions. NEC and Anthropic will co-develop vertical AI products for finance, manufacturing, and local government, sectors that demand high security, Japanese regulatory compliance, and quality standards that offshore AI vendors often struggle to meet. The first wave will use Claude Cowork (Anthropic's business-user desktop agent) combined with NEC's operational know-how in each vertical. Anthropic's models will also feed into NEC's cybersecurity services, sharpening its SOC (security operations center) offerings.

Pillar 2: BluStellar integration. BluStellar is NEC's in-house "value creation" DX platform. Claude Opus 4.7 (Anthropic's flagship model) and Claude Code (Anthropic's coding agent) will be integrated into two BluStellar scenarios first, one focused on data-driven executive decision-making, the other on customer understanding, before expanding across the platform.

Pillar 3: 30,000-person rollout. Every employee across the NEC Group, roughly 30,000 people, will get Claude. NEC is calling this the largest AI-native engineer workforce at any Japanese company. Anthropic will provide technical support and training, and NEC will stand up an internal Center of Excellence dedicated to AI talent development.

Paul Smith, Anthropic's Chief Commercial Officer, framed the deal in long-term language: "A significant step in a long-term effort to build the future of AI in Japan together." NEC Executive Vice President and COO Toshifumi Yoshizaki echoed the framing, calling it a partnership meant to "play a central role in transforming Japanese customers through AI." Both sides are signaling this is not a pilot, it's a bet.

NEC's AI Strategy Is Actually Three Layers

To understand what this partnership means, you have to see that NEC's AI strategy already has three layers running in parallel.

Layer 1, In-house LLM "cotomi." NEC released its first 13-billion-parameter Japanese-specialized LLM in July 2023, followed by cotomi Pro and cotomi Light in 2024. These models aren't meant to beat GPT-5 or Claude at general intelligence. They're designed to be compact, Japanese-optimized, and deployable on-premises, ideal for clients handling confidential data or resistant to cloud-based AI. This is NEC's "narrow and deep" specialist layer.

Layer 2, Anthropic's Claude (this new deal). Where cotomi is narrow-and-deep, Claude is broad-and-deep. It covers complex reasoning, long-context processing, software development, and multimodal tasks that small specialist LLMs simply can't handle. Claude doesn't replace cotomi; it fills the gaps cotomi was never designed to fill.

Layer 3, The national "Japan AI Foundation Model Development" company. Established in April 2026 together with SoftBank, Honda, and Sony, this joint venture is targeting a 1-trillion-parameter domestic foundation model focused on "Physical AI", AI that controls robots and factory equipment. It's backed by roughly $6.3 billion in Japanese government funding over five years. This is the "long horizon, sovereign AI" layer.

In other words, NEC is hedging across three time horizons simultaneously: short-term (deploy Claude now to generate revenue), medium-term (use cotomi for Japanese-language and on-prem cases), long-term (build sovereign Physical AI over a decade). It's neither total dependence on Silicon Valley nor total isolation in domestic tech, a portfolio approach that neither Hitachi nor NTT has replicated as clearly.

Why Claude, Not ChatGPT?

When you line up Japan's enterprise AI deals, the assignments are striking:

  • NTT Data × OpenAI (May 2025): First ChatGPT Enterprise reseller in Japan, accelerator program for 100 major corporations
  • Fujitsu × Cohere (July 2024): Equity stake in the Canadian AI startup, joint Japanese LLM development
  • NEC × Anthropic (April 2026): First Japanese "global partner," Claude Cowork for vertical solutions

Why did NEC pick Anthropic over OpenAI? Three reasons stand out.

First, market momentum has flipped. In Menlo Ventures' 2025 enterprise survey of around 150 technical leaders, Anthropic's share of the enterprise LLM API market reached 32%, edging past OpenAI's 25% for the first time. Whatever you think of general-purpose chatbot popularity, Claude is winning corporate deployments.

Second, Anthropic's "safety and controllability" brand fits regulated industries. Anthropic was founded on an AI Safety research mission, and that reputation resonates with compliance-heavy sectors, finance, government, healthcare. NEC's first-wave targets (finance, manufacturing, local government) are exactly where that fit matters most.

Third, the product stack aligns. Claude Code is a leading AI coding agent for developers, while Claude Cowork is the desktop agent for non-developer business users. Together, they cover both halves of NEC's 30,000-person workforce, engineers and office staff, in a single product family. OpenAI has competing products, but Anthropic's stack apparently matched NEC's "developer + operations" use case better.

Notably, Hitachi Systems announced a separate (but smaller) Anthropic deal on the same day. The "global partner" title, however, went only to NEC.

The Market at Stake: $43 Billion by 2029

These mega-deals aren't happening in a vacuum. Japan's enterprise AI market is expanding at a rate that dwarfs most of the developed world.

According to IDC Japan's March 2026 forecast, domestic AI spending will grow from ¥2.37 trillion (~$14.8 billion) in 2025 to ¥6.89 trillion (~$43 billion) in 2029, roughly triple in four years, at a 36% CAGR. By 2029, AI will account for 20% of total Japanese IT spending.

Japan's AI infrastructure investment alone is projected at ¥821 billion (~$5.1 billion) in 2026, up 18% year-over-year. By 2028, AI infrastructure spending will surpass non-AI infrastructure spending, a structural tipping point that means AI has graduated from "experimental deployment" to "core social infrastructure."

Against this backdrop, NEC's Anthropic bet is a land grab. Winning early credibility in finance, manufacturing, and government verticals now positions NEC to capture a disproportionate slice of that growth.

The Geopolitics: Japan Hedges Toward the US

There's also a geopolitical subtext no Japanese enterprise IT deal can escape.

Frontier foundation models are, practically speaking, a US–China duopoly: OpenAI, Google, Anthropic, Meta, and xAI on one side; DeepSeek, Alibaba, and Baidu on the other. Japanese firms can develop their own models, but to deliver practical AI services to enterprise customers today, they must partner with one of the frontier labs.

Every major Japanese systems integrator has planted its flag on the American side of the board. NTT Data went with OpenAI. Fujitsu bet on Canada's Cohere. NEC just bet on Anthropic. No major Japanese IT vendor has productized a Chinese foundation model for its enterprise customers. Between economic security concerns, data handling rules, and customer-trust considerations, the practical option set has narrowed to US AI.

At the same time, the Japanese government is pouring roughly ¥1 trillion into the SoftBank-led "Japan AI Foundation Model Development" company to eventually reduce dependence on US AI, especially for Physical AI use cases tied to Japan's manufacturing base.

NEC's new partnership navigates this tension elegantly: use American AI for the next three to five years of customer demand, while co-building sovereign Japanese AI for the decade after. Call it the "dual-track" strategy. It may become the default template for Japanese enterprise AI.

The Risks: Claude Dependency Isn't Free

Not everyone is bullish. The partnership carries real risks that rarely make it into press releases.

First, Anthropic-side geopolitical risk. The US continues to debate AI export controls and model licensing. If future US administrations impose restrictions on foreign access to frontier models, Japanese mission-critical systems deeply integrated with Claude could face disruption. This is not theoretical, semiconductor export controls have shown how fast Washington can move.

Second, cost opacity. Claude Opus 4.7 delivers top-tier performance, but it's not cheap. At 30,000-person scale, total cost of ownership comparisons versus on-premises alternatives like cotomi will eventually become a board-level question.

Third, data sovereignty. Claude Cowork runs on user desktops with access to local files. Reconciling that with Japan's Personal Information Protection Act, Financial Services Agency guidelines, and sector-specific rules is not a trivial engineering problem. It's a compliance marathon that will determine how deep Claude can actually embed in Japanese enterprises.

NEC acknowledges these constraints and emphasizes "safe and trustworthy AI implementation", but so does every competitor making similar deals. The real differentiator will be execution: whether 30,000 NEC employees can actually convert Claude access into customer value, at scale, across regulated industries.

What's It Like in Your Country?

In the US, the enterprise center of gravity is visibly shifting from OpenAI toward Anthropic. In China, the government is pushing state buyers toward domestic models like DeepSeek and Qwen. In Europe, the EU AI Act has introduced a slower, more cautious corporate adoption curve.

What about your country? Which foundation model providers, American, Chinese, local, are your biggest companies actually deploying? Is your government investing in sovereign AI infrastructure? And how comfortable are highly regulated sectors like banking and public administration with installing desktop-level AI agents from overseas vendors?

Whether NEC's Anthropic bet becomes the winning template for Japanese enterprise AI or deepens Japan's dependence on US AI is genuinely unclear. The answer won't be visible until those 30,000 NEC engineers actually start shipping customer solutions in late 2026.

References