Usage-based GPU access
Planned hourly GPU access for inference, model serving, rendering and batch processing. Start with the workload, GPU memory and usage window you need.
DISTRIBUTED GPU INFRASTRUCTURE
GPU compute.
Built to scale.
ETERNApod is developing a distributed GPU cloud: modular pods at sites with suitable power and fibre, connected through one shared software layer. Planned access for inference, rendering and AI workloads—without buying your own GPU infrastructure.
For investors & strategic partnersIn development · First milestone: prototype validation
01
02
03
+
+
MODULARITY / CONCEPT VISUALConceptual illustration of the distributed pod architecture.
01 / THE INITIATIVEPOWER × FIBRE × AI
01 / VISION
Compute power.
Built where it matters.
Built to grow across locations.
ETERNApod aims to use existing power and fibre at suitable residential, industrial and educational sites through modular GPU pods. The proposed architecture connects pods across locations into one centrally managed service network.
START SMALL. VALIDATE. SCALE.From a prototype to a pod fleet, then modular facilities and a long-term campus vision.
FOR AI TEAMS & ENTERPRISES
The compute you need.
A model that fits.
Tell us what you want to run.
We will assess a suitable pilot configuration.
FLEXIBLE WORKLOADS
Reserved capacity
A planned route for teams with predictable demand: discuss an agreed capacity allocation, schedule and commercial term.
Enterprise pilots
Scope a defined use case before a longer commitment. Review hardware fit, data handling, performance criteria and integration needs together.
Pre-launch enquiries. Pricing, availability and service terms will be confirmed after technical review and prototype validation.
Discuss a pilotFOR SITE OWNERS
Your location.
Part of the network.
The proposed host model turns suitable space, power and fibre into a location for a centrally operated GPU pod.
Propose a locationA suitable site
Space, power capacity, fibre connectivity and physical access, subject to a site assessment.
The pod and software layer
The proposed operator model covers hardware operation, central orchestration, and power and internet costs.
Fixed host rent
The host receives agreed fixed rent under a hosting contract. Terms depend on site suitability and deployment approval.
FOR INVESTORS & STRATEGIC PARTNERS
Understand the business.
Follow the milestones.
A phased infrastructure model: deploy repeatable GPU units, connect capacity through shared software, and expand as technical validation and customer demand support the next step.
Compute consumption
Planned GPU usage fees, reserved capacity agreements, AI usage credits and storage services.
One managed network
Pods at qualified host sites. Central management coordinates allocation and operations across locations.
Evidence before expansion
Prototype testing, paid pilots, utilisation and operating costs inform each fleet and facility decision.
What the first milestone must prove
- Technical performance under a 72-hour load test
- Site readiness, measured power draw and noise
- Pilot customer fit and commercial demand
- Repeatable installation and operating economics
Development-stage initiative. Fleet and campus figures are targets. Investment discussions are subject to due diligence and agreed terms.
Discuss the opportunity02 / OUR APPROACH
The parts behind
the whole.
One location. One pod. A shared network.
Explore the three components of the model.
PROPOSED SERVICE MODEL
GPU compute
Model serving, batch processing, rendering and suitable fine-tuning workloads for AI teams. Planned options include hourly GPU usage, reserved capacity, AI usage credits and storage.
Services are part of the development roadmap.03 / HOW THE NETWORK WORKS
Distributed locations.
One coordinated system.
Workload
Compute requirements from AI teams and enterprise users.
Portal + API
A planned shared access layer for capacity requests, usage tracking and billing.
Orchestration
Software designed to route workloads to suitable pods and manage capacity and continuity.
Pod fleet
Modular GPU units across locations, operating under shared management.
This is the proposed architecture. Network and service performance will be validated through prototype and fleet testing.
04 / THE FIRST POD
Compact footprint.
Focused capacity.
Target configuration for Phase 1.
An engineering design to validate through testing.
RTX PRO 6000 Blackwell
Combined GPU memory
Target peak power draw
Steel cabinet form factor
96 TB raw storage + NVMe cacheDX cooling design≤50 dB noise target
Target values. Final hardware specifications depend on procurement and prototype testing. Raw storage differs from usable capacity.
05 / SCALING ROADMAP
From one pod
to what comes next.
Each stage validates the foundations for the next. The figures below represent phased development targets, not operational capacity.
Automatic phase tour · 2 seconds per stage
FIRST VALIDATION MILESTONE
1POD
Prototype & field validation
Validate the first pod through a 72-hour load test, power consumption and noise measurements, installation checks and pilot use cases.
Priority: a repeatable, measurable first unit.
06 / WHO IT IS FOR
AI builders.
Space providers.
AI teams & enterprises
Flexible capacity for model serving, inference, rendering, batch processing and suitable fine-tuning applications.
Site owners
A fixed-rent hosting model for sites with suitable power, fibre and physical access. The operator is expected to cover power and internet costs.
Strategic partners & investors
Opportunities to participate in hardware, energy, locations and infrastructure as the development programme moves from prototype to scale.
07 / YOUR QUESTIONS
A clear perspective.
What stage is ETERNApod at today?
The initiative is in development. The first objective is prototype production and technical validation. Capacity, production timelines and growth figures on this site are targets, not an operational fleet or service guarantees.
Can a pod be installed at any location?
No. Power capacity, fibre connectivity, physical access, cooling conditions and required permissions must be assessed for each site. Hosting arrangements follow this suitability review.
How long do production and installation take?
The target is a 45-day standard cycle, or 30 days with all components in stock. These timelines depend on component availability and supply agreements; they are not guaranteed delivery times. Critical components that are out of stock can significantly extend the schedule.
Is the distributed network suitable for every AI workload?
The focus is inference, model serving, batch jobs and suitable fine-tuning applications. Due to latency between locations, large model training requiring tightly synchronised operation across thousands of GPUs is outside the initial distributed model’s focus.
Who is behind the initiative?
ETERNApod is being developed with the support of HANSconstruction and ETERNA+ Partners. HANSconstruction brings construction and delivery experience, while ETERNA+ Partners contributes its architectural and design approach to this infrastructure vision.
08 / A NOTE FROM OUR CEO
Thinking ahead.
Building responsibly.
“A bold vision earns trust by proving itself in small steps. With ETERNApod, our ambition is to bring engineering, design and sound management together to develop scalable AI infrastructure. Validate first, then grow. For me, that is where lasting value begins.”
Veli Tunahan Bay
CEO
HANSconstruction · ETERNA+ Partners
Bringing together a background in civil engineering, architecture and construction management, Veli Tunahan Bay has led HANSconstruction and ETERNA+ Partners since 2019.
Education: London Metropolitan University · Architecture, 2022–2025.
STRATEGIC ADVISOR
Ahmet Bozoğlu
Strategic & Capital Markets Advisor
Founder & CEO, FinSPAX LLC (USA)
25+ years of U.S. capital markets experience, specializing in capital raising strategy, investor relations, SPAC transactions and U.S. public market access.
Education: Long Island University, 2000–2004.
09 / CONTACT
Let’s take the next step
together.
Discuss a pilot, a host partnership or an investment opportunity.
Istanbul / Maslak · London / Mayfair, W1J 6BD