More work completed
in the factory.
Prefabrication moves assembly and testing into a controlled setting while site works proceed in parallel. NVIDIA applies this approach to an integrated AI system.
Sports venues. New possibilities.MODULAR AI INFRASTRUCTURE
Big nights.
Bigger potential.
AI computing. Available power. Useful heat.
Engineered around the life of your venue.
GE HyperSport · A Griffiths Evans companyExplore the concept ↓
01 / The opportunity
Stadiums. Arenas.
Event campuses.
Your venue’s
next act.
Sports venues are built for extraordinary peaks. GE HyperSport brings engineering expertise to the infrastructure between them.
We review your electrical capacity, connectivity and heat demand, then design the infrastructure for modular AI computing around your venue’s event calendar. We bring the venue and compute operator together through a coordinated engineering design.
01
Make capacity work harder
Establish what is available, when, and on what terms.
02
Explore a new revenue stream
Assess hosting partnerships against costs and utilisation.
03
Put heat to use
Match recovered heat to demand, at the right temperature.
02 / Inside the moduleHigh-density compute.
High-density compute.
A complete engineering system.
Select a numbered callout to explore the module and see what GE designs and verifies around it.
Module concept · Power, compute and cooling
Seven racks.
One coordinated system.
NVIDIA DSX POD is the basis of design for GE HyperSport. GE designs the power, cooling, connectivity and heat-recovery infrastructure needed to deploy the pod at your venue.
7Compute racks per pod
72Rubin GPUs per NVL72 rack
504GPUs across seven racks · 7 × 72
800 VDCElectrical input architecture
GE develops the electrical load schedule, cooling duty and site layout around the NVIDIA DSX POD architecture. NVIDIA DSX POD: 21 October 2026 · NVL72: 16 March 2026. View sources ↗
03 / Beyond the computeUseful heat.
Useful heat.
By design.
Capture heat from computing.
Connect it to your venue.
We design heat-recovery connections for water preheating, low-temperature heating and heat networks. Our assessment brings together heat output, demand, pipe routes, heat pumps and commercial value.
See GE’s integration scope ↗
01 Heat exchange
02 Pumping & controls
03 Heat rejection
Cooling system concept
Up to 45°C
NVIDIA’s stated Vera Rubin coolant temperature. GE designs the heat-recovery circuit and delivery temperature around the heating system.
A complete heat balance
GE checks what can be recovered, when it can be used, and how surplus heat is rejected safely.
04 / The Griffiths Evans roleFrom available power
From available power
to a deployable project.
We review your infrastructure and design the elements needed to connect, protect and operate the modules. Specialist partners provide the compute platform and its operation.
01Power infrastructure
We review grid agreements, metered event and non-event demand, transformers, switchgear, protection and resilience. We design the upstream distribution, conversion, cabling, earthing and metering needed to connect the NVIDIA DSX POD.
Deliverable
A verified capacity envelope and a coordinated electrical design.
02Compute & connectivity
We define physical and MEP interfaces, coordinate fibre routes and carrier requirements, and work with the specialist compute provider on service access, equipment loads and operating requirements.
Deliverable
A coordinated installation, with clear responsibility for the IT platform.
03Liquid cooling
We establish thermal duty, flow rates and temperature limits with the supplier. We design facility pipework, pumps, heat exchangers, isolation, leak detection and controls, and coordinate coolant distribution units.
Deliverable
Cooling designed for the actual module duty and local conditions.
04Heat rejection
We size and locate dry coolers and any supplementary cooling. We review ambient temperatures, noise, airflow recirculation, plant access and resilience with the venue and specialist suppliers.
Deliverable
A dependable cooling route even when recovered heat is not required.
05Site integration
We plan the services layout, maintenance access and connection routes, and coordinate foundations, delivery logistics, drainage, fire strategy and security with the relevant specialists.
Deliverable
A module that fits the venue’s physical and operational constraints.
06Heat recovery
We compare available heat with hourly and seasonal demand. We assess heat exchangers, thermal storage and heat pumps where needed, then design the connection to suitable venue or neighbouring heating systems.
Deliverable
A quantified heat-use case, with temperatures, operating hours and economics stated.
07Controls & event protection
We define venue-priority operating limits, electrical interlocks, BMS interfaces, alarms and response sequences. With the compute operator, we establish which workloads can reduce demand, pause or move when event loads increase.
Deliverable
An operating strategy that protects the venue and reflects the compute service agreement.
08Testing & commissioning
We prepare interface specifications and acceptance criteria, review supplier submissions and coordinate integrated tests for normal operation, peak demand, power interruption, cooling faults and heat-recovery bypass.
Deliverable
Documented evidence that the installed interfaces operate as intended, together with an operational digital twin and remote monitoring.
05 / The benefits
More from your infrastructure.
More work completed
Months to days
at a prepared site.
NVIDIA describes deployment in as little as 14 days at a prepared site. GE coordinates the power, cooling, connectivity and site works that make the installation ready.
More output from
available power.
NVIDIA DSX MaxLPS manages AI workloads within an available power envelope. GE defines the venue’s electrical limits, metering and controls interfaces.
Less cooling-water
consumption.
Closed-loop liquid cooling circulates coolant through the equipment and external heat exchangers. GE selects the heat-rejection plant for the site’s climate and cooling duty.
Flexible demand.
Events come first.
GE designs venue-priority controls and agrees the compute operating strategy with the operator, aligning electrical demand with the event calendar.
New income.
Useful recovered heat.
GE develops the business case for hosting income and recovered heat, bringing together investment, electricity costs, operating hours and heating demand.
Explore the NVIDIA architecture and technical publications behind our basis of design. Technical references ↗
Start with the venue.
Build the evidence.
01
Discover
Review power data, space, connectivity, heat demand and the event calendar.
02
Define
Agree a capacity envelope, operating model and supplier configuration.
03
Design
Coordinate electrical, cooling, controls and site interfaces with the delivery team.
04
Prove
Specify integrated tests and verify the installation against agreed criteria.
Let’s explore what’s possibleYour venue.
↗Register your interest ↗Your venue.
The next opportunity.
Tell us about your venue, your infrastructure and your ambitions. We’ll help establish your available power and the next steps.
Technical references +
NVIDIA DSX POD and Vera Rubin architecture form the basis of design for GE HyperSport. The illustrations show the GE HyperSport venue concept.
- NVIDIA DSX POD announcement ↗21 October 2026
- DSX POD technical architecture ↗21 October 2026
- DSX power, cooling and operation ↗21 October 2026
- NVIDIA video: AI factories in modular containers ↗GTC Berlin · October 2026
- GTC Berlin keynote transcript — seven-rack configuration ↗21 October 2026
- Vera Rubin NVL72 rack specification ↗16 March 2026