Road 2 Web Summit 2023
Recognized in Lisbon for a model that opens the AI infrastructure economy to a wider range of people — a small entry ticket to AI infrastructure, presented as digital and financial inclusion.
NODE UNION
A platform for participation in global AI infrastructure.
Four GPU packages · Entry from $25 · Monthly scenarios from 15%1
3 sites across three countries
$25 RTX entry package
$500 NVL72 package
485 → 950 TWh data center demand2
Vendor-neutral architecture: configurations are selected for the workload, availability, and economics.
NVIDIA BLACKWELLtraining and inference
AMD INSTINCTalternative GPU pools
HGX / GB300dense AI systems
NVMe + OBJECTdata storage tiers
PRIVATE AIisolated environments
Stargate NorwayNarvik, Norway · base GPU pool
Verne GlobalKeflavík, Iceland · modular AI Factory
LUMIKajaani, Finland · HPC environment
T1 EnergyMo i Rana, Helgeland · 50 MW allocatedTraining, fine-tuning, scientific compute, and batch inference · 50–150+ MW planning range.
Parque Industrial OlmosEmpalme Olmos, Canelones · 20–40 MW utility capacityLatin American hub for training, fine-tuning, scientific compute, and batch inference.
SAKURA internetIshikari Bay New Port, Hokkaido · modular expansion10–20 MW planning modules with staged growth toward 100+ MW.
Explore a participation scenario.
This demonstration model illustrates potential operating cash flow over different terms. It does not promise a fixed return or replace the terms of a specific pool.1
$
$25$500
Base scenario over 12 months1
$22,50
USD · modeled operating cash flow before taxes
Conservative$28.35
Base$40.50
High Utilization$48.60
Term 1–36 months
Modeled utilization 45–85%
Base scenario over 24 months ≈ $45
Compare Packages
The final configuration requires site-specific commercial pricing: GPU allocation TBD · performance TBD · power requirements TBD.
Important. Results depend on utilization, rental rates, energy costs, equipment availability, reserves, and operating expenses. They may be lower than modeled, zero, or negative.
Before activation, the agreement, pool composition, fees, reserves, exit terms, and calculation methodology are disclosed.
From capital to hardware.
From utilization to cash flow.
1Capital Pooling
Participant packages form a shared pool for hardware and the infrastructure module.
2Phased Deployment
Capacity is purchased and commissioned as demand and supply are confirmed.
3External Capacity Rental
Compute capacity is provided through GPU Cloud, IaaS, bare metal, and dedicated clusters.
4Operating Cash Flow
Revenue covers energy, networking, cooling, service, insurance, and reserves.
Package from $25 → infrastructure pool → deployment → external capacity rental → costs and reserves → potential distribution or reinvestment.
Initial Deployment
3 sites.
Three countries.
A distributed network reduces dependence on a single power system, jurisdiction, or point of connectivity.
Energy
Resource-aware energy.
No blanket claims.
Priorities include solar, wind, and safe local generation, closed-loop liquid cooling, and transparent water and energy accounting.
Transparency
Terms in plain view.
Beyond the interface.
The site, calculation methodology, agreement, and exit process should be available before a decision is made.
Scale
Modular AI Factory.
Grow module by module.
Repeatable megawatt-scale modules standardize power, cooling, networking, security, and monitoring. Local resources determine each site's limits.
Capacity built for real workloads.
From API inference to private environments and large-scale training clusters.
Training and Reasoning
Blackwell and HGX clusters, plus alternative GPU pools for large models.
Serverless Inference
On-demand GPUs and APIs for real-time request processing.
Private AI
Isolated environments for banking, healthcare, industry, and sensitive data.
Science and Engineering
Simulation, large-scale data analysis, and HPC workloads.
Bare Metal and Racks
Dedicated servers, private segments, and long-term clusters.
Generative Services
Text, image, video, voice, and multimodal applications.
Transparency for every pool.
Participants should see the underlying assumptions, operating metrics, and legal terms—not a promise.
1
Capacity and Availability
Installed configuration, uptime, utilization, and workload type are disclosed for every pool.
2
Economics
Average rental rates, energy, networking, cooling, service, and operating expenses are reported separately.
3
Reserves and Upgrades
Insurance, maintenance reserves, and the hardware technology-refresh schedule are accounted for separately.
4
Legal Structure and KYC
The legal structure, parties' rights, identity checks, timelines, and exit process are available before participation.
Infrastructure locations and edge facilities.
A unified view of the initial deployment, reference infrastructure, and candidate edge facilities represented on the map.
Stargate Norway · live
Stargate NorwayNarvik, Norway · base GPU pool
AI is becoming physical infrastructure.
485 → 950 TWh
The IEA estimates that data center electricity consumption could nearly double between 2025 and 2030.2
≈ 3%of global electricity by 20302
3 sitesinitial distributed deployment
≈ 1 MWrepeatable AI Factory module
Awards and international recognition.
Ten awards, shortlists, and finalist places between 2023 and 2026 — from Web Summit and MWC Barcelona to the UN SDG Action Awards and the Zayed Sustainability Prize.
GSMA Global Mobile Awards (GLOMO) 2024
Shortlisted for Best Mobile Innovation for Emerging Markets. The GLOMO Awards honour the companies, individuals, and governments driving innovation across the mobile ecosystem.
SET Tech Festival 2024
Included in the SET100 List 2024 — the top 100 emerging companies in clean energy and climate technology, selected from more than 430 applicants across over 75 countries.
UN SDG Action Awards 2024
Finalist in the Impact category. UNDP Administrator Achim Steiner described the finalists as pioneers who tackle the root causes of injustice and the planetary crisis.
Africa Tech Festival Awards 2024
Winner of the Innovation for Impact Award for digital infrastructure that gives clients a practical route to growth, scalability, and cost-efficiency across Africa.
DCD Global Awards 2024
Finalist for Middle East & Africa Data Centre Project of the Year — a category for projects that push the boundaries of data centre design, engineering, and construction.
Zayed Sustainability Prize 2025
Named a finalist for an energy-efficient AI data centre architecture. The UAE’s global sustainability award selected 33 finalists for 2025 from almost 6,000 submissions.
LEAP 2025 — Rocket Fuel Pitch Competition
Top 10 with a distributed financing model for AI compute capacity. Rocket Fuel brings technology startups from around the world to pitch before leading investors and industry experts.
World Summit Awards Global Congress 2025
WSA Winner in the Inclusion & Empowerment category for advancing digital inclusion in AI infrastructure — a category for solutions that bridge the digital divide and expand access.
Datacloud Global Awards 2026
Shortlisted for the Data Centre Innovative Energy Solution Award for a responsible approach to working with the communities around AI data centres.
Demonstration block. Award names, categories, and organizers are shown as published by the award programs; images are provided by the organizers and event photography.
Choose a participation format.
Compare each package by entry amount and monthly scenario rate, then review the selected configuration in the calculator.
NVIDIA Light
PopularModeled monthly return
15%
Non-compounded scenario rate
+$45net result · 12 months
Entry amount$25
Choose a Package
14% of spots remaining
NVIDIA Pro
ProfessionalModeled monthly return
17.5%
Non-compounded scenario rate
+$105net result · 12 months
Entry amount$50
Choose a Package
AMD Pro
PerformanceModeled monthly return
20%
Non-compounded scenario rate
+$600net result · 12 months
Entry amount$250
Choose a Package
NVIDIA Advanced AI
Sold outModeled monthly return
25%
Non-compounded scenario rate
+$1,500net result · 12 months
Entry amount$500
Ask a Manager
Sold out · ask a manager
The final configuration depends on availability and economics: NVIDIA Blackwell / Blackwell Ultra, HGX and GB300 NVL72, AMD Instinct MI350 / MI355X, high-speed networking, NVMe cache, object storage, and private segments.
What to review before participating.
Three categories of materials to review before making an infrastructure participation decision.
Contracting parties, scope of participation, jurisdiction, KYC, payments, and the participant's position relative to other parties.
Utilization, rental rates, energy, fees, operating expenses, insurance, and the technology-refresh reserve.
Exit process, residual value, currency and regulatory risks, supply, downtime, and technology obsolescence.
Frequently asked questions.
Key details about the model. Legally binding terms are defined by the documents for each specific pool.
What does the $25 NVIDIA Light package include?
NVIDIA Light is the entry level for pooled RTX 5090 / 4090 capacity—not the price of a standalone GPU or server. The exact allocation, legal structure, availability, and terms are specified in the documents for the selected pool.
How does capital become active compute capacity?
Funds are pooled and allocated to hardware procurement, deployment, and maintenance. The resulting capacity is then provided to external clients through GPU Cloud, IaaS, bare metal, dedicated clusters, and inference services.
How is the estimated result calculated?
Energy, connectivity, maintenance, fees, downtime, and technology-refresh reserves are deducted from cash receipts based on actual utilization and client pricing. The remaining cash flow may be available for distribution or reinvestment.
Are distributions guaranteed?
No. Results depend on actual cash flow and may be lower than estimated, zero, or negative. The interface presents demonstration scenarios—not a fixed return.
What hardware does the platform support?
The model is vendor-neutral. Configurations may use NVIDIA Blackwell Ultra, HGX and GB300, AMD Instinct MI350/MI355X, high-speed networking, NVMe and object storage, and private AI environments.
Why are three sites used?
Operating across three countries diversifies operational risk and allows the platform to account for the availability of energy, connectivity, engineering resources, and client requirements. Every pool receives separate reporting.
What are the main risks?
Capacity utilization, energy and connectivity costs, currency fluctuations, construction and delivery timelines, technical failures, financing, regulation, competition, and hardware obsolescence.
Who is the platform designed for?
The platform is intended for a global audience, with Nigeria identified as one of its initial development markets. Product availability and the contractual model depend on local regulation and identity-verification requirements.
Participate in global
AI infrastructure.
Three sites, pooled hardware capacity, and transparent reporting for modeled scenarios.