When a Japanese tech giant wants in on the AI boom, it usually picks a side. Fujitsu just refused to.

On a single Wednesday in May 2026, the 90-year-old company announced deals with two rival AI labs at once — Anthropic and OpenAI — pairing the makers of Claude and ChatGPT in the same press cycle. Its rival NEC, when it tied up with Anthropic a month earlier, deliberately chose one. Fujitsu's "why not both" is not indecision. It is a bet on a very particular idea of how corporate Japan will actually buy AI.

Two deals, announced the same day

On May 27, 2026, Fujitsu unveiled a strategic partnership with Anthropic and, separately, the start of a collaboration with OpenAI. The wording matters: the Anthropic deal is framed as the deeper of the two — Fujitsu becomes an Anthropic "strategic partner" — while the OpenAI tie-up is described as a collaboration getting underway.

With Anthropic, Fujitsu gets early access to Anthropic's newest Claude models and will combine them with its own tools: the Takane large language model and the Fujitsu Kozuchi AI platform. Roughly 100,000 group employees will use Claude in daily work and software development. With OpenAI, Fujitsu's staff will lean on ChatGPT Enterprise and the Codex coding agent, with a stated focus on manufacturing clients, where Fujitsu already has a strong customer base.

Both deals point in the same direction. Fujitsu wants to use the tools internally first — it calls itself "Customer Zero" — then sell what it learns. And both feed into cybersecurity work, including cooperation with the Japanese government on defending critical infrastructure. President Takahito Tokita, in a statement on the announcement, said pairing Fujitsu's accumulated industry know-how with Anthropic's frontier AI should help generate new value across sectors and move toward an AI-driven society that people can rely on with confidence.

Image announcing Fujitsu's strategic partnership with Anthropic

Source: Fujitsu press release

NEC picked one. Fujitsu picked two.

The contrast with NEC is the real story here. In April 2026, NEC became Anthropic's first Japanese "global partner" — a designation only one company in Japan holds — and rolled Claude out to about 30,000 employees. NEC weighed Anthropic against OpenAI and chose Anthropic, full stop. (We dug into why in our earlier piece on the NEC–Anthropic deal.) Hitachi signed its own Anthropic agreement around the same time.

Fujitsu's badge is different. It is a "strategic partner," not a "global partner" — and it did not stop at one lab. Where NEC made a single, deliberate bet, Fujitsu hedged. That difference is philosophical. NEC's logic is alignment: pick the partner whose product stack best fits its customers and commit. Fujitsu's logic is neutrality: stay model-agnostic, keep more than one frontier supplier on the bench, and let each customer's needs decide which engine goes under the hood.

It is worth noting that Fujitsu's own Takane model was co-developed with the Canadian AI startup Cohere. Counting Cohere, Anthropic, and OpenAI, Fujitsu now works with three foreign AI shops. For a company selling itself as a neutral integrator, that is the point, not a contradiction.

What a "Forward Deployed Engineer" actually does

Both Fujitsu deals revolve around a term that may be unfamiliar outside the enterprise-software world: the Forward Deployed Engineer, or FDE. The model was popularized by the US data-analytics company Palantir, and it works like this. Instead of shipping software and hoping the customer figures it out, the vendor sends an engineer to sit inside the client's business — a bank, a factory, a city office — to learn how the work really happens, design the AI use case on the spot, build it, and stay until it sticks.

This is what Fujitsu is actually selling. Not Claude. Not ChatGPT. The integration. The AI lab supplies the engine; Fujitsu supplies the mechanic who installs it, tunes it for a Japanese regulatory environment, and makes sure it survives contact with a real workplace. Seen that way, having two engines in the garage is an asset, not a hedge gone soft.

Fujitsu isn't investing in AI. It's reselling it.

Here is the distinction that gets lost in the headlines. When Microsoft "partnered" with OpenAI, it put in something on the order of 13 billion dollars and wired OpenAI into its cloud. When Google backed Anthropic, it became a major shareholder. SoftBank pours tens of billions into OpenAI and its Stargate data-center project. Those are capital relationships — equity, cloud lock-in, ownership stakes.

Fujitsu's deals carry no announced equity. Neither did NEC's. These are go-to-market partnerships: the Japanese company gets early model access, training, and the right to build and resell solutions, while the AI lab gets a distribution channel into corporate Japan's banks, manufacturers, and ministries. Money is not changing hands the way it does in a Silicon Valley mega-investment.

That leaves three distinct postures among Japan's tech heavyweights. SoftBank is the investor, betting capital on the AI infrastructure layer. NTT is the builder, developing its own homegrown model, tsuzumi, now in government trials. And Fujitsu and NEC are the integrators — the ones who take someone else's model and carry it the last mile into the enterprise.

Not Microsoft-OpenAI. Not Google-Anthropic.

So when you read "Fujitsu partners with OpenAI," resist the instinct to picture the Microsoft relationship. The shape is different and the stakes are smaller. Japan's enterprise software market still runs heavily through system integrators. A regional bank or a government ministry does not buy AI off a website and wire it in over a weekend; it commissions a project from Fujitsu, NEC, or NTT Data, complete with compliance review, support contracts, and someone to call when it breaks.

That structure makes the relationship symbiotic rather than acquisitive. The AI labs need those channels to reach customers who would never sign a direct contract with a four-year-old startup in San Francisco. The integrators need frontier models they have no realistic path to building themselves. Neither side has to own the other for the deal to work.

What it says about Japan's AI strategy

Step back and a familiar dual-track pattern appears. In the short term, corporate Japan is leaning on US frontier models — Claude, ChatGPT — because they can deliver value now. In the longer term, it is funding sovereign alternatives: Fujitsu's Takane, NEC's cotomi, NTT's tsuzumi, and a national foundation-model company. Reliance today, independence tomorrow.

Fujitsu's two-lab twist adds one more layer to that hedge. If one AI provider stumbles, raises prices, or changes its terms, the other can carry the customer. It also keeps Fujitsu from being captured by any single vendor's roadmap. The risk, of course, is the obvious one: running two frontier platforms plus your own model is complex, and critics are right to ask whether a partnership announcement is the same thing as a working product on a customer's desk. That gap is where this story will actually be decided over the next year.

What is already clear is the philosophy. Japan's biggest tech companies are buying into AI through partnerships rather than acquisitions — and Fujitsu just doubled down by partnering with two rivals on the same day.

In your country, when a big bank or a government office adopts AI, does it go straight to OpenAI or Anthropic — or does it still run through a large integrator first? We would like to hear how AI actually reaches the enterprise where you are.

References