The Votranex Token Factory

From power to tokens,
on one accountable stack.

Votranex runs a Token Factory: AI data centers, model serving and API, all on one stack. Every token you buy from us is made in the region where you use it, on infrastructure we operate or have certified.

Start building →Talk to us
770MW+in the Votranex pipeline
2MW–300MWmodular nodes to campuses
In-regionyour data stays local
How it works

The Votranex Token Factory

One accountable stack, from power to tokens.

  1. Power & sitesSecured grid and renewable power
  2. AI data centers2MW nodes to 300MW campuses
  3. AI computeReserved, heat-dense clusters
  4. Model servingOptimised open models
  5. Token APITokens, made in-region

From power and sites to the API, we take responsibility for every layer, so every token we deliver can be traced back to where it was made.

AI ClustersReserved floor space: GPUs, power and data center capacity
Dedicated InferenceA reserved production line: reserved serving capacity
Token APIOutput: tokens, paid per use
Products

One factory. Three ways to buy.

Reserve a whole cluster, rent a dedicated production line, or pay per token.

Reserved floor space

AI Clusters

Compliant, reserved GPU clusters with power, from 2MW modular nodes to 300MW campuses, for teams that build and serve AI models.

Learn more
Reserved production line

Dedicated Inference

Your models, fine-tuned models or private deployments on reserved capacity. Predictable speed, no shared queues.

Learn more
Pay per token

Token API

Leading open models, billed per token. Works with the OpenAI and Anthropic API formats. Made in-region, with your data kept local.

Learn more

All three run on the Votranex Token Factory, one stack from power to API.See our infrastructure →

Infrastructure · AI Data Centers

The foundation of our Token Factory.

We bring purpose-built AI data centers to market, from 2MW modular nodes to 300MW campuses, close to where people use AI, with power secured and designed for both serving and training models.

770MW+
in the Votranex pipeline
Sized to demand.Modular sites of a few megawatts for fast local inference, and large campuses for model builders.
Built for heat-dense GPUs.Liquid cooling and high-efficiency design.
Demand-backed.Our Token API and Dedicated Inference are the first tenants of every data center we build.

The Votranex pipeline: from 2MW modular nodes to 300MW campuses, power secured, delivered in stages.

Phase 1 built15MWBuilt200MWPlanned

Ishikari Campus

Hokkaido, Japan
15MW + 200MW planned
100MW from 2028100MW2028200MWby 2030

Kitakyushu Campus

Kyushu, Japan
300MW
50MW from 202950MW2029200MWby 2032

Ibaraki Campus

Kanto, Japan
250MW
Delivering from 20283–5 nodes · 2–4MW each2027 → 2028

Modular Renewable Nodes

Across Japan
2–4MW per site
Partner with us
Landowners and power holders, construction partners, co-investors.
Develop a data center with us
About

Built by operators who have done it before.

The AI era will be won by those who can turn power into tokens: reliably, locally and responsibly. Doing that takes three things rarely found in one team. You need to know how to build and scale a technology company, how to develop and finance large infrastructure, and how to bring capital and partners together across continents.

Our founding team brings all three. Between us, we have built a company from a blank page to an exit, closed cross-border infrastructure and real-estate deals, and backed technology companies on both sides of the Pacific.

1company built from zero to exit
US$100Mraised as founders
300+people led across five offices
US$800Mdeployed as investors
Contact

Let's build your token supply.

Tell us what you're building, whether that's reserved capacity, a private deployment or a data center partnership. Our team will get back to you within five business days.

Prefer email? hello@votranex.ai

AI Clusters

Compliant, reserved GPU clusters with power, cooling and operations included, for companies that train and serve AI models and need capacity close to their users.

  • From 2MW modular nodes to 300MW campuses, delivered in stages as your demand grows.
  • Faster than building new. Sites come with power secured, so you skip the multi-year wait for a grid connection.
  • Fully managed or bare-metal, with optional model serving and operations from our inference team.
  • Compliant by design. Customer screening, export-control checks and data-residency rules are built into every contract.
  • Serve through us (optional). Make your models available to local customers through our Token API.
For: AI model companies, AI labs, cloud GPU providers, and governments and institutions building their own AI capacity.

Capacity pipeline: 770MW+

SiteRegionScaleFirst power
Ishikari CampusHokkaido, Japan15MW Phase 1 · 200MW expansion plannedPhase 1 built
Kitakyushu CampusKyushu, Japan300MW100MW in 2028
Ibaraki CampusKanto, Japan250MW50MW in 2029
Modular Renewable NodesAcross Japan2–4MW per siteFrom 2028

Reserve early: capacity is allocated in the order of committed demand.

Reserve capacitySee all sites
Dedicated Inference

Private endpoints on reserved GPUs: open models, your fine-tuned versions or your own model, with speed and throughput guaranteed in your contract.

  • Reserved, not shared. Capacity set aside for you, and isolated per customer on request.
  • Bring or build your model. We fine-tune open models for your field and language, and can deploy them in your private environment.
  • Enterprise-ready. Single sign-on, audit logs, private network connections, data processing agreements, and compliance documents for each region.
  • Pricing models: per hour, per month, or a committed token volume.
For: enterprises in regulated industries, SaaS platforms, video and media platforms, and companies that need control over their data.
Talk to us
Token API

One API key for DeepSeek, Kimi, MiniMax, GLM, Qwen, gpt-oss and more, plus leading open models for image and video generation. It works with your existing OpenAI and Anthropic code, and is served in-region on infrastructure we operate or have certified.

Why Votranex

  • Factory-direct pricing. We run the hardware and serving ourselves, so you pay close to production cost rather than reseller markups.
  • Your data stays in-region. Prompts, files and outputs are processed and stored only in the region you choose. We never use them for training.
  • Built for heavy workloads. High throughput and low delay for demanding work such as coding agents, AI video production, multi-step agent workflows and processing documents in bulk.
  • New models on day one. We work directly with model makers, so new releases go live quickly.
  • Switch between models freely. Route, fall back between models and balance load from one key.
For: developers, AI-native product teams, creative and media studios, and companies launching products in new markets.
Start building
Hokkaido, Japan · Phase 1 built

An AI data center campus in Ishikari, Hokkaido. The 15MW Phase 1 is built, and a further 200MW expansion is being planned.

15MWPhase 1 · built
Up to 200MWPhase 2 · in planning
  • Why this site. Hokkaido's cool climate supports efficient cooling, and the region has strong access to renewable energy.
Reserve capacity
Kyushu, Japan · 300MW

A 300MW AI data center campus in Kitakyushu, Fukuoka Prefecture, delivered in phases.

100MWPower from 2028
+200MWBy 2030
  • Why this site. Large-scale power for western Japan, close to major demand centers in Asia.
Reserve capacity
Kanto, Japan · 250MW

A 250MW AI data center campus in Ibaraki Prefecture, within reach of the Greater Tokyo area.

50MWPower from 2029
+200MWExpected by 2032
  • Why this site. Large-scale capacity close to Japan's largest market, without inner-Tokyo grid waits of up to 10 years.
Reserve capacity
Across Japan · 2–4MW per site

Modular AI data center nodes built from shipping containers, placed at existing renewable energy plants operated by a leading Japanese renewable energy operator.

2–4MWPer node
6–8 monthsConstruction to delivery
3–5 sitesIn development 2027, delivering 2028
  • Why this model. Running on renewable power from day one, with fast delivery. Nodes sit close to local users for low-delay inference.
Reserve capacityDevelop a data center with us

Cookie Settings

Choose which cookies we may use. You can change this at any time.

Strictly necessaryNeeded to run the website, remember your choices and protect it from spam and abuse.
Always on
PerformanceAnonymous, aggregated page-view statistics that help us improve the website. No cookies or personal identifiers.