explainx.ai0k
TrendingAI News TodayPathwaysSkills
Pricing
explainx.ai

Upskill in AI — 16 free pathways, live workshops & bootcamps, and 50+ courses from practitioners. Plus the skills, tools, and MCP servers to practice on.

follow us

follow on google

Add explainx.ai as a preferred source

corporate training

support@explainx.ai

get started

Find your pathTake Free Evaluation

community

Join the community

learn

mind: share how you thinkpathways — start freeworkshopsbootcampscoursescompare Explainxcertificationsmock testsexplainx universitycorporate traininglearn skills & mcp

discover

skillsmcp serversexplainx mcptoolsmdx readeragentsllmsdesignsdictionarypeopleagi trackerfelony benchranks

company

aboutvisionmissionteaminstructorsteach on explainxpartnershipscommunityhackathonscareers

content

daily AI newsstate of AI — live resultsblogreleasespromptsgeneratorsresource libraryfor LLMsexplainx.ai kids

solutions

all solutionsdeveloper upskillingmarketing upskillingproduct manager upskillingleadership upskilling

newsletter · weekly

Get AI news, tools, and insights in your inbox.

supportcontactprivacytermsdata rightshow we create contentsubmission guidelines

© 2026 AISOLO Technologies Pvt Ltd

explainx.ai

On this page

  • TL;DR: the reported facts
  • What exactly was reported?
  • Why 500 MW is a big number
  • Why Osaka, and why NTT Data?
  • What this means if you build with Claude
  • Where it fits in the compute race
  • Local politics and power: the risks
  • What is claimed, and what is verified
  • Bottom line
  • Related reading
← Back to blog

explainx / blog

NTT Data Reportedly Close to Building Anthropic's First Japan Data Center: Up to 500 MW in Osaka

Anthropic, AI Infrastructure, Data Centers, Japan, NTT

Part of Anthropic and Claude

Nikkei reports NTT Data is in final talks to build Anthropic's first Japan data center in Osaka: up to 500 MW, 1 trillion yen. What's reported and what isn't.

Oct 13, 2026·8 min read·Yash Thakker
add explainx.ai
go deep
NTT Data Reportedly Close to Building Anthropic's First Japan Data Center: Up to 500 MW in Osaka

Nikkei Asia reports that NTT Data Group is in the final stages of talks to build a data center in Osaka for Anthropic, which would be the Claude maker's first dedicated data center in Japan. According to Reuters, which relayed the story, the facility could draw up to 500 megawatts of power and cost as much as 1 trillion yen, roughly $6.3 billion, with partial operations starting as early as 2028. Nothing is signed, and neither company has confirmed the plan.

This post separates what the two reports actually say from what is inference, then lays out why a project of this size matters for people who build with Claude, for Japan's power grid, and for the wider race to secure AI compute that explainx.ai has been tracking, including Anthropic's reported $517B in compute commitments.

TL;DR: the reported facts

table · 2 cols
QuestionAnswer (per Nikkei and Reuters)
Who is building it?NTT Data Group, a unit of telecom giant NTT
Who is the customer?Anthropic
Where?Osaka, western Japan
How big?Power intake of up to 500 MW, among the largest in Japan
How much?Up to 1 trillion yen (about $6.3 billion)
When?Partial operations as early as 2028; agreement targeted by end of fiscal 2026 (March 31, 2027)
Is it confirmed?No. Based on multiple unnamed sources; NTT declined to comment; no Anthropic comment in either report

What exactly was reported?

The original report is by Nikkei Asia reporters in Tokyo and Osaka, published on October 13, 2026 Japan time. Per Nikkei Asia, NTT Data Group is in the final stage of talks to build the facility, citing multiple sources. The headline claim is that this would be Anthropic's first dedicated data center in Japan, with a build cost that could reach $6.3 billion and a partial opening expected by 2028.

The Reuters write-up adds the numbers that make the scale concrete: up to 1 trillion yen, up to 500 MW of power intake, and a target of signing by the end of the fiscal year that closes on March 31. It also records that NTT declined to comment and that the NTT Group could not be reached immediately.

Two details are worth keeping straight. First, the 500 MW figure is a power ceiling, not a count of chips. Second, the cost is described as "up to," so the final figure may be lower or change once an agreement is drafted. Treat both as negotiating-stage numbers.

Why 500 MW is a big number

For scale, a campus of up to 500 MW is in the range that the industry now calls a gigawatt-class build in spirit, even if it sits below one gigawatt. Sites of that size draw power the way a mid-sized city does, which is why they are planned years ahead and often tied to dedicated generation.

NTT's own recent Osaka project is much smaller. NTT Global Data Centers' Osaka North Data Center in Ibaraki, Osaka Prefecture, is planned at 36 MW of IT power across two buildings (18 MW in phase one, with service starting in the second half of fiscal 2027, and another 18 MW by 2032), per its newsroom. Note that IT load and power intake are different measures, but an Anthropic campus of up to 500 MW would still be an order of magnitude beyond that project.

That gap explains why the power question dominates the story. Securing 500 MW from the grid takes time, and Nikkei's page does not state the power figure at all; it comes from the Reuters relay.

Why Osaka, and why NTT Data?

Osaka is already Japan's second hub for cloud and colocation, built partly as a disaster-resilience counterweight to Tokyo. Western Japan has land and grid headroom that the Tokyo region increasingly lacks, and a facility there can still serve users across the country with low latency.

NTT Data is a natural builder: its NTT Global Data Centers arm already operates and is expanding colocation capacity in the Osaka area.

Neither source says who would supply or operate the chips inside the building, or which cloud arrangement would sit on top. That matters: Anthropic has trained and served Claude across several chip platforms and cloud partners, and a Japan facility could be a hosted build where NTT provides shell, power and cooling while Anthropic or a partner brings the hardware. The reports do not say, so we will not guess. For the background on how Anthropic buys capacity, see our explainer on the Nvidia contracted value of over $180B with Anthropic.

Data center power grid link: transformer connected to a server cabinet representing the power demand of a 500 MW AI data centerData center power grid link: transformer connected to a server cabinet representing the power demand of a 500 MW AI data center

What this means if you build with Claude

If you use Claude from Japan, nothing changes today. On the reported timeline, the first capacity would arrive in 2028 at the earliest, and Anthropic has not announced a Japan region or an in-country processing product.

Still, there are three practical things to watch:

  1. Data residency. Japanese enterprises and public-sector buyers often require that data stay in the country. A local facility is the usual precondition for offering that. Anthropic's Tokyo presence has been growing, and our coverage of Code with Claude Tokyo shows the developer side of that push.
  2. Latency. Serving models closer to users cuts round trips, which helps interactive agents and voice features more than batch jobs.
  3. Capacity and rate limits. Every new block of compute is one more reason limits can loosen over time. That connects directly to current-generation models like Claude Opus 5.5, whose usage is bounded by available capacity.

None of these is a promise. They are the standard reasons a lab builds regionally, offered here as likely motives, not as announced features.

Where it fits in the compute race

The Osaka plan, if it closes, is one more data point in a year when AI labs have turned into some of the largest infrastructure buyers on earth. We covered the scale of Anthropic's obligations in the $517B compute commitments explainer, and the geopolitical side in China's 24 GW versus the US's 56 GW, where the lesson was that gigawatts overstate effective compute when chips, not power, are the binding constraint. The same caution applies here: 500 MW of power is not 500 MW of useful AI hardware until the accelerators arrive.

It also lands in the same week as reports that Tencent is leasing 100,000 AI chips from Oracle in Southeast Asia, a reminder that Asia is where much of the next wave of capacity is being arranged, by very different players and under very different export-rule constraints.

Partnership puzzle pieces joining, representing the reported NTT Data and Anthropic agreement still being negotiatedPartnership puzzle pieces joining, representing the reported NTT Data and Anthropic agreement still being negotiated

Local politics and power: the risks

Big data center projects have run into resistance in many places, from siting disputes to water and grid concerns. The United States offers a live example in San Francisco's data center moratorium. Japan has generally been more welcoming to data center investment, but a 500 MW load is large enough to need grid-operator coordination, local permitting and, plausibly, new generation, any of which can slip the 2028 start.

Other risks sit on the deal itself. Talks can stall over price, power guarantees or hardware supply. The fiscal-year signing target is an aim, not a commitment. And the cost ceiling can shift with exchange rates, since the yen figure and the dollar figure move apart over time.

What is claimed, and what is verified

table · 2 cols
ClaimStatus
NTT Data is in final talks to build an Anthropic data center in OsakaReported by Nikkei, relayed by Reuters; unconfirmed by either company
Up to 500 MWReported by Reuters citing Nikkei
Up to 1 trillion yen (about $6.3B)Reported by Nikkei and Reuters
Partial operations from 2028Reported by both
Agreement by March 31, 2027Reported by Reuters as the target
Who supplies the chipsNot reported

To check for updates, watch Anthropic's newsroom, NTT Group and NTT Data press releases, and Nikkei's follow-ups. A signed agreement would normally be announced by both sides at once.

Bottom line

If the report holds, Anthropic would move from renting capacity in other regions to anchoring a purpose-built campus in one of the world's most important AI markets, with a builder that already knows Osaka. The numbers are large but unconfirmed, the timeline is long, and the biggest open questions are power and hardware. For Claude users, this is a signal about where capacity is headed, not a feature you can use. We will update this post when either company comments.

Weekly digest3.6k readers

Catch up on AI

Curated AI updates on agents, skills, and MCP — delivered to your inbox. Unsubscribe anytime.

Related reading

  • Anthropic's Reported $517B Compute Commitments, Explained
  • Nvidia–Anthropic $180B Contracted Value: What Builders Pay
  • China Has 24 GW of Data Center Capacity vs 56 GW in the US
  • Tencent Reportedly Leases 100,000 AI Chips From Oracle
  • San Francisco Data Center Moratorium
  • Code with Claude Tokyo
  • Claude Opus 5.5 launch, benchmarks and pricing

Details are accurate as of October 13, 2026, and rest on Nikkei Asia and Reuters reporting from unnamed sources; the deal has not been confirmed by the companies.

Spotted something out of date? Let us know.
Yash Thakker

Written by

Yash Thakker

Yash is an AI expert with over 300K learners. Join his workshops →

View Yash Thakker in People in AI →

Related posts

Oct 12, 2026

China's "$3.9 Trillion Grid Buildout" for AI: What Is Claimed and What Is Verified

A claim about a $3.9 trillion Chinese grid buildout to turn wasted power into an AI edge is circulating. The best-documented figure is State Grid's 4 trillion yuan plan for 2026 to 2030, roughly $574 billion. This post separates the claim from the record and explains why grid capacity matters for AI.

Oct 12, 2026

FERC Rejects PJM Curtail-First Data Center Plan as Amazon Preps for Entergy Cuts

A viral headline says US regulators blocked a proposal to cut power to AI data centers. The verifiable event is FERC's October 9 order rejecting most of PJM's large-load plan, which would have let new data centers connect first and be curtailed in emergencies. Here is what is confirmed, what rests on anonymous sources, and why it matters.

Oct 12, 2026

NVIDIA Opens Vera Rubin Rack Designs at OCP: MGX 1.2, DSX Software and Grid Specs

At the OCP Global Summit, NVIDIA contributed Vera Rubin rack and compute-tray designs, MGX 1.2, a Vera CPU rack spec, open DSX software and grid-resilience standards. Here is what each piece is and what it means for AI factory builders.