While the news items were flowing individually, looking at them side-by-side revealed a different landscape.
In just two months between April and May 2026, the AI partnership map for Japan's major System Integrators (SIers) was completely redrawn.
Fujitsu, Hitachi, NEC, NTT Data, and NRI.
By laying out which camp these five companies joined, what they intend to do, and what comes next, several things have become clear.
What you will gain from this article:
- The differences and strategic intentions of AI partnerships among major Japanese SIers
- The true meaning behind the "difference in terminology" when Fujitsu partnered with two companies on the same day
- The structural dilemma facing NTT Data and what will happen next
- The gambling dynamic of the "First Move x Single AI" strategy by NEC and Hitachi
- What Anthropic and OpenAI will do next and the competitive risks to SIers
- Three actions AI promotion leads can take starting today
*This article is written based on public information and primary sources (press releases, FSA official announcements). It includes some of my personal analysis (as of May 27, 2026).
Chapter 1: What Happened in Two Months
To be honest, I didn't expect announcements to come in such rapid succession over such a short period.
Let's organize what happened between April and May 2026 chronologically.
On April 23, 2026, NEC announced a strategic partnership as Anthropic's "first global partner in Japan," declaring it would deploy Claude to approximately 30,000 employees.
On May 18-19, Hitachi announced a strategic partnership with Anthropic, launching Claude deployment for 290,000 employees and the integration of Frontier AI into "Lumada 3.0."
Then, on May 27, Fujitsu announced partnerships with both Anthropic and OpenAI on the same day, releasing two press releases at once.
Around this time, other significant movements occurred at the government and financial infrastructure level, including the launch of Anthropic's "Project Glasswing" (April 7), the establishment of a public-private working group led by the Financial Services Agency (May 14), and the granting of Claude Mythos access to Japan's three megabanks.
Why did it concentrate in these two months?
It has been about six months since Anthropic opened its Tokyo office in October 2025 and appointed Hidetoshi Tojo (formerly of Snowflake Japan) as Head of Japan. After building the foundation, they moved to finalize agreements all at once. Additionally, May was a time when competitive pressure to "secure SIers in the Japanese market first" reached its peak, as both Anthropic and OpenAI launched FDE-type JVs with PE firms in the US.
Chapter 2: The Five Major SIers—Who Joined Which Camp?
Let's compare them equally.
- Fujitsu (FY2025 Revenue: 3.55 trillion yen)
- Anthropic: "Strategic Partnership." Company-wide deployment to 100,000 people, providing a 1,000-person engineer team to customers.
- OpenAI: "Collaboration." Utilizing ChatGPT Enterprise and Codex for FDE business.
- Positioning: The SIer among the five that most clearly打ち出した a multi-AI route.
- Hitachi (FY2025 Revenue: ~9.78 trillion yen)
- Anthropic: "Strategic Partnership." Claude deployment to 290,000 people, establishing Frontier AI Deployment Centers in North America, Europe, and Asia (initially 100 people → 300).
- OpenAI: No official partnership announcement.
- Positioning: Anthropic-only. Specialized in Physical AI and OT (Energy, Transportation, Power).
- NEC (FY2025 Revenue: ~3.42 trillion yen)
- Anthropic: "First Global Partner in Japan." Claude deployment to 30,000 people, specialized for finance, manufacturing, and local governments.
- OpenAI: No official partnership announcement.
- Positioning: Anthropic-only. Differentiation focused on regulated sectors.
- NTT Data (FY2025 Revenue: 4.63 trillion yen)
- Anthropic: No official partnership (as of May 27, 2026). However, there are circumstances mentioned later.
- OpenAI: "Global Strategic Partnership." Japan's first and the world's second (after PwC) ChatGPT Enterprise reseller. Target of 100 billion yen cumulative scale by the end of FY2027.
- Positioning: Currently the standard-bearer for the OpenAI camp. However, the situation is changing.
- NRI (Nomura Research Institute) (FY2024 Revenue: 736.5 billion yen)
- Anthropic: "Japan's first Anthropic Certified Reseller" (via Amazon Bedrock, Nov 6, 2025). Claude for Enterprise already introduced internally, full-stack implementation SI system.
- OpenAI: Handles ChatGPT Enterprise (listed as an official OpenAI Japan partner).
- Positioning: A balanced approach handling both, with Anthropic as the axis. Rather than large-scale company-wide deployment like Fujitsu, they position themselves as an implementation partner closer to finance and consulting.
Chapter 3: The True Meaning of Fujitsu Partnering with Two Companies on the Same Day
This is my analysis, prefaced with "this is how it can be read."
Reading Fujitsu's press releases, the first thing I noticed was the difference in terminology.
Contract with Anthropic: "Strategic Partnership"
Contract with OpenAI: "Collaboration"
I believe this difference is not accidental.
Generally, the expression "Strategic Partnership" carries a stronger nuance of long-term, comprehensive cooperation than mere technical use or sales cooperation. The announcement with Anthropic specifically mentioned "company-wide deployment to 100,000 people," "formation of a 1,000-person engineer team," and a "focus on cybersecurity and critical infrastructure."
On the other hand, the announcement with OpenAI disclosed no details regarding personnel, amounts, or duration.
Despite releasing both on the same day, it appears to the reader that they are being ranked differently.
Read this way, Fujitsu's message is: while taking a positioning as an "SIer that can master multi-AI," their strategic center of gravity is placed on Anthropic for critical infrastructure, government agencies, and defense.
The FDE model Fujitsu has been working on since 2023 overlaps significantly with the design philosophy of the $1.5 billion JV Anthropic launched with Blackstone and Goldman Sachs in May 2026. It's a method of "stationing engineers at customer sites to redesign operations." The "Strategic Partnership" with Anthropic functions as the implementation foundation for Fujitsu to deploy this FDE model in the Japanese market.
As a side note, in Fujitsu's English press release for the OpenAI announcement, there was an error where it still said "Anthropic's advanced AI models." This is a trace of both announcements being prepared in parallel.
A single term can reveal a company's center of gravity.
Chapter 4: NTT Data's Structural Dilemma
Actually, this is the most interesting part.
On April 24, 2025, NTT Data announced a strategic partnership with OpenAI as Japan's first ChatGPT Enterprise reseller. It was a serious commitment with a target of "100 billion yen cumulative by the end of FY2027."
On the other hand, they have no official partnership with Anthropic at this time.
However, looking at the participant list for the working group of the "Public-Private Liaison Meeting on Strengthening Cybersecurity Measures in the Financial Sector Against AI Threats" established by the FSA on May 14, 2026 (from the official FSA press release), NTT Data is sitting at the same table as Anthropic.
Fujitsu, Hitachi, NEC, NRI, Anthropic Japan, OpenAI Japan, and NTT Data are all lined up together.
In other words, NTT Data has an asymmetrical relationship: a "100 billion yen reseller contract" with OpenAI, while being a "fellow participant in a government critical infrastructure defense working group" with Anthropic.
This structure could lead to the OpenAI contract becoming a drag.
ChatGPT Enterprise and Claude for Enterprise are direct competitors fighting for the same budget. If NTT Data officially becomes an Anthropic reseller or partner, pressure will arise from the OpenAI side asking "which one are you prioritizing?" Having publicly stated a 100 billion yen goal, they also have accountability for that figure.
However, participation in government working groups is unrelated to reseller contracts.
I predict NTT Data's next realistic move will be a compromise: "partnering with Anthropic for technical cooperation and implementation without becoming a reseller." The same structure where Fujitsu differentiated between "Strategic Partnership (Anthropic) + Collaboration (OpenAI)" could happen to NTT Data.
As cyber defense networks based on Glasswing proceed with government backing, the state of being "uncontracted with Anthropic" will become a competitive weakness in various scenarios.
It is particularly serious in the financial sector. NTT Data is an SIer with overwhelming strength in systems for financial institutions. In the future, financial institutions will increasingly ask, "How will you build a vulnerability detection and disclosure system based on Mythos-class AI?" If the responsible SIer has "no official relationship with Anthropic," it is, simply put, an accountability issue. NTT Data is currently in a difficult position to answer the question, "Can you really protect us?"
To speak without reservation, for NTT Data, this isn't a "price paid for partnering with OpenAI," but rather an unexpected development where "Anthropic deeply embedded itself into Japan's critical infrastructure after they partnered with OpenAI." The decision to set a 100 billion yen goal wasn't a mistake. The map just changed too fast.
Chapter 5: NEC and Hitachi Took the Lead. But it's a "Single AI" Gamble
Reading this far, NEC and Hitachi might look like the "star students" who partnered with Anthropic first.
It is true they made the first move.
NEC partnered on April 23 as Anthropic's "first global partner in Japan." Hitachi announced Claude integration into "Lumada 3.0" and deployment to 290,000 people on May 19. Since Fujitsu partnered with both on May 27, NEC and Hitachi moved 1 to 5 weeks earlier.
These two companies were the first to create a track record of large-scale partnerships as Anthropic solidified its footing in Japan.
However, this first move contains a gamble.
Both companies have partnered only with Anthropic and have no official relationship with OpenAI. In other words, they have made a "Single AI Model" choice.
This contrasts with Fujitsu's explicit "Multi-AI route."
If a customer chooses Claude, they are strong. But when a customer says they want to use ChatGPT, NTT Data will come to the negotiation as the "SIer partnered with OpenAI." How will NEC and Hitachi respond then? Will they try to persuade them to "please use Anthropic," or will they partner with OpenAI later?
At this point, there is no answer.
In my view, NEC and Hitachi chose Single AI not because they were "confused," but because they "chose." NEC, in particular, is concentrated in regulated sectors like finance, manufacturing, and local government, which aligns well with Anthropic's safety and reliability pitch. Hitachi has the vertical domain of OT (Operational Technology), where its own physical assets serve as a differentiator.
The choice of Single AI might look like a weakness, but for these two companies, it has the potential to function as "selection and concentration."
Chapter 6: The Real Competitor is Not the AI Company Itself, but the "AI Implementation Force"
I believe this is something many people haven't noticed yet.
On May 4, 2026, Anthropic established a $1.5 billion enterprise AI services company with Blackstone, Goldman Sachs, and Hellman & Friedman. OpenAI also raised $4 billion from 19 PE investors including TPG and Bain Capital to establish "The Deployment Company" on the same day.
Both companies have adopted the FDE (Forward Deployed Engineer) model popularized by Palantir. This means Anthropic and OpenAI have set up systems to enter the "AI implementation and business transformation consulting" space that SIers have traditionally occupied.
Fortune described this as "Anthropic moving to compete directly with the world's largest consulting firms like McKinsey."
For Japanese SIers, this movement presents two distinct problems.
[Problem 1: The risk of Japanese SIers' FDE declarations becoming "pie in the sky"]
The essence of the FDE model is to get deeply involved in the customer's business outcomes rather than just "selling man-months." While billing may not become entirely performance-based, the evaluation axis changes from traditional phase-based, man-month SI.
Based on my 23 years of observing large corporate environments, the resistance to changing SI contract structures is far higher than any technical barrier. The practice of stacking man-months for each phase—"requirements definition → design → development → testing"—provides a sense of security for the client, as they can easily see "what they are paying for and where." Changing the evaluation axis brings the difficult problem of "how to define results" first. The reality is that many client companies are not prepared to face this challenge.
I read the current rush of FDE declarations from Fujitsu, NEC, and Hitachi as a defensive move to control that discussion under SIer leadership before Anthropic and OpenAI move in directly.
[Problem 2: Why it will take time for "Black Ship" AI implementation companies to fully enter Japan]
I don't think the FDE-type JVs of Anthropic and OpenAI will move into the Japanese market immediately. However, the reason is separate from the contract model.
The US JVs are designed based on Blackstone's portfolio of over $1 trillion. The circuit to deploy AI implementation using "investor status" into a group of companies controlled by PE (Private Equity) does not exist in Japan.
However, there is another circuit: the "government-sanctioned" model of the FSA working group. It is highly likely to spread from megabanks to critical infrastructure operators to public agencies and ministries. Once a foothold is established in this circuit, there is a high possibility that Anthropic or OpenAI will establish JVs involving capital with Japanese companies.
And Accenture is already moving.
In December 2025, Accenture signed a three-year strategic partnership with Anthropic and launched an independent organization called the "Accenture Anthropic Business Group." They are conducting Claude training for 30,000 employees and co-developing FDE-type offerings for CIOs.
While domestic SIers are at the stage of declaring they "will do FDE," Accenture already has an implementation system of 30,000 people. When going to propose "FDE-type AI implementation support" to a large company's CIO, Accenture comes to the meeting as a "partner with a proven track record of a three-year contract with Anthropic."
This is a quite dire situation.
It's more accurate to read the FDE declarations of Fujitsu, NEC, and Hitachi as a countermeasure against Accenture as much as a defense against Anthropic and OpenAI.
If Anthropic establishes a JV involving capital with major Japanese financial institutions, trading houses, or PE funds, it will be the signal that "the positioning of SIers has fundamentally changed." It's not happening right now, but we can't take our eyes off it.
Chapter 7: Three Actions AI Promotion Leads Can Take Starting Today
I've written a long analysis, but here is what you should do as a field AI promotion lead:
- Confirm which camp your primary SIer has joined this week. Who is your main SIer: Fujitsu, NEC, Hitachi, NTT Data, or NRI? Check whether that SIer is partnered with Anthropic, OpenAI, or both, and re-check if there is any impact on the proposals for ongoing AI projects. Projects where the SIer effectively decides the model selection will increase in the future.
- Create a one-page draft of a "Multi-AI Policy." Fujitsu's move to set a "premise of using multiple AIs" is a signal for AI introduction in large corporations. If implementation proceeds without internal criteria for using Claude vs. ChatGPT (data sensitivity, use case, cost), it will become uncontrollable later. Just verbalizing your current judgment criteria on a single sheet of A4 will allow you to face vendor proposals proactively.
- Regularly track the progress of Glasswing. The FSA working group is for financial institutions, but the same movement will spread to public sectors, energy, transportation, medical, and defense. According to Anthropic's official announcement, Claude Mythos discovered over 10,000 high-to-critical vulnerabilities across partners within just one month of Project Glasswing's launch. Start confirming now whether your system and SIer contracts include provisions for handling "AI-driven vulnerability detection and disclosure."
From outside a large corporation, "partnerships between SIers and AI companies" might seem like a distant topic.
But the era where the SIer building your internal systems designs them on the premise of using Claude is just around the corner.
To speak plainly, the worst thing for an AI promotion lead is to remain passive, thinking "I can just leave it to the SIer."
I want you to be in a position where you can choose the models the vendors select.
That is all.
Finally, let me ask you one thing.
Is your company's main SIer in the Anthropic camp, the OpenAI camp, or have you not yet grasped the situation?
Thank you for reading this far.
As a "window-seat manager" at a large corporation, I continue to experiment by throwing Claude Cowork into the field every day.
For managers in the same position who are struggling with "wanting to use Claude but not being able to use it internally" or "not knowing which AI to use and how," I share actual working examples and prompts on X from time to time.
→ @madogiwacowork
[Correction/Addendum: May 28, 2026]
A reader pointed this out: Hitachi, Ltd. also signed an MoU for a strategic partnership with OpenAI on October 2, 2025. The content primarily focuses on providing power, transmission/distribution equipment, and air conditioning technology for AI data centers, which is a different layer from the "Claude deployment to 290,000 people and Lumada integration" with Anthropic. However, my description regarding Hitachi's relationship with both companies was inaccurate. Thank you for the correction.
