05Projects

AI NEOCloud, in two phases.

Two deployments that bring high-performance GPU compute on demand to enterprises, AI developers, research teams and cloud partners.

Project ITP
1.2 MW · Q4 2026
Project PX5
4.5 MW · Q3 2027
Model
GPU-as-a-Service

05.1Roadmap

From live hubs to 4.5 MW.

TodayHubs A & B active
Q4 2026Project ITP go-live
Q3 2027Project PX5 go-live

01Project ITP

Go-live Q4 2026

AI NEOCloud & tokenization platform

Build a specialised AI cloud platform that delivers high-performance GPU compute on demand, with tokenized service credits and capacity access to make AI infrastructure easier to consume, allocate and commercialise.

Compute
AI training, inference & HPC — reserved or usage-based
Service credits
Redeemable for GPU hours and cloud services, with entitlement, usage and settlement tracking

Capacity

Hardware

72 NVIDIA B300 servers

Business model

  • Recurring GPU-as-a-Service revenue
  • Reserved capacity contracts
  • Storage & platform services
  • Tokenized service-credit ecosystem
  • 01

    GPU NEOCloud

    Compute on demand for the workloads that matter.

    • On-demand GPU compute for AI training, inference and HPC
    • Flexible access through reserved capacity or usage-based services
    • Supporting infrastructure, storage and workload orchestration
  • 02

    Tokenization

    Digital service credits that make capacity easy to allocate.

    • Digital service credits redeemable for GPU hours and cloud services
    • Track capacity entitlements, usage and settlement
    • Potential future token-based financing or investment models, subject to regulatory review
  • 03

    Deployment & value

    Enterprise-grade from day one.

    • 72 NVIDIA B300 servers within a 1.2 MW data-centre environment
    • Designed for scalable, enterprise-grade AI workloads
    • Target customers: AI developers, enterprises, research teams and cloud partners

Any future token-based financing or investment models are subject to regulatory review.

02Project PX5

Go-live Q3 2027

AI NEOCloud comprehensive phase

Expand high-performance AI cloud capacity through a 4.5 MW data-centre deployment powered by NVIDIA GB300 NVL72 systems, delivering scalable GPU infrastructure for demanding enterprise and next-generation AI workloads.

Workloads
Model development, fine-tuning, inference & accelerated computing
Growth
Power, cooling, networking and operations aligned for phased expansion

Capacity

Hardware

NVIDIA GB300 NVL72 rack-scale systems

Business model

  • GPU-as-a-Service
  • Reserved capacity contracts
  • AI infrastructure partnerships
  • Managed compute services
  • 01

    Next-gen GPU cloud

    Rack-scale Blackwell for the largest models.

    • NVIDIA GB300 NVL72 rack-scale AI infrastructure
    • On-demand GPU compute for AI training and inference
    • High-performance environment for large-scale AI models and HPC workloads
  • 02

    4.5 MW deployment

    Built for density, designed to grow.

    • Expanded data-centre capacity designed for dense AI compute
    • Power, cooling, networking and operations aligned to rack-scale systems
    • Scalable architecture to support phased growth and service expansion
  • 03

    Customer value

    Capacity that matches demand.

    • Flexible GPU capacity for enterprises, AI developers and research teams
    • Supports model development, fine-tuning, inference and accelerated computing
    • Enterprise-focused service delivery with capacity options to match demand

Each cell represents 0.1 MW, on one shared scale for both phases.

05.2Side by side

Two phases, one platform.

 Project ITPProject PX5
PhaseAI NEOCloud & tokenization platformAI NEOCloud comprehensive phase
Go-liveQ4 2026Q3 2027
Capacity1.2 MW4.5 MW
Hardware72 NVIDIA B300 serversNVIDIA GB300 NVL72 rack-scale systems
FocusGPU compute on demand, with tokenized service creditsScalable capacity for large-scale models and HPC
Business modelGPU-as-a-Service · reserved capacity · storage & platform services · service creditsGPU-as-a-Service · reserved capacity · AI infrastructure partnerships · managed compute
CustomersAI developers, enterprises, research teams and cloud partners