
ZXO W1
ConceptSINGLE GPU / MAX-Q
A 96 GB Max-Q GPU target, with a 300 W GPU power budget.
Explore W1ZXO / AI WORKSTATIONS & SERVERS
Considered design. Purposeful compute.
Explore a focused collection of system concepts for professional work and enterprise AI projects in the United States.
Design concepts and target configurations.
Engineering validation pending. Not available to order.

NEW / ZXO GB10 DESKTOP AI CONCEPT
A compact AI desktop direction with generous cooling space, an open airflow architecture and a carefully finished aluminum enclosure. Explore the design and the proposed GB10 platform.
Concept renders. Hardware integration and performance validation pending. Availability has not been announced.

NEW / ZXO GB300 FLAGSHIP AI CONCEPT
A considered tower design for a proposed GB300 desktop AI platform. Explore the metalwork, eight exterior views and a flagship configuration target with 32 TB of raw NVMe storage.
Exterior concept and engineering targets. OEM integration, storage qualification and availability remain to be confirmed.

NEW / ZXO B300 ROSSO EXTERIOR CONCEPT
Explore the red-and-black ROSSO server study, twelve exterior views and NVIDIA DGX B300 reference specifications for training, inference and AI reasoning.
Discover ZXO B300New: B300 high-density rack / 12 views
Exterior concept imagery. Hardware integration, final configuration and engineering validation remain to be confirmed.

ZXO 5090 / GRAPHICS CARDS
Explore the ZXO 5090 graphics-card collection: sculptural red ROSSO COUTURE and graphite-and-champagne TRUNK ATELIER. Two exterior directions, together in one place.
Explore all 5090 designsAlso explore TRUNK ATELIER — 20 black-and-champagne views
Exterior concept imagery. GPU-board integration, dimensions, power and cooling require engineering validation.

01 / WORKSTATIONS
Three GPU targets for creative applications, technical work and local AI. Start with the software you use and the work you need to do.

SINGLE GPU / MAX-Q
A 96 GB Max-Q GPU target, with a 300 W GPU power budget.
Explore W1
SINGLE GPU / WORKSTATION EDITION
A flagship single-GPU target with 96 GB of ECC graphics memory.
Explore W2
DUAL GPU / WORKSTATION EDITION
A two-GPU target with a separate 96 GB of memory on each card.
Explore W3Max-Q and Workstation Edition are distinct GPU variants. Final system configurations, power and thermal requirements remain subject to engineering validation.
Compare workstations02 / AI SERVERS
Rack-server concepts for larger projects. RTX PRO Server Edition, H200 NVL and HGX H200 represent different integration paths.

2U TARGET / RTX PRO SERVER EDITION
A two-GPU server concept for enterprise AI and professional compute.
Explore S2
4U TARGET / H200 NVL
A four-GPU H200 NVL concept targeting a supported PCIe platform.
Explore S4
8U-CLASS TARGET / HGX H200
An eight-GPU HGX H200 platform target with H200 SXM GPUs.
Explore S8Rack heights are exterior design targets. Platform sourcing, chassis fit, cooling, electrical requirements and final configurations are not yet validated.
Compare server targets03 / OUR APPROACH
ZXO is developing its own collection of professional workstation and AI server concepts. Our direction brings restrained industrial design together with workload-led platform planning.
Applications, model size, data, project scale and deployment location help define a useful starting point.
Discuss the GPU variant, memory needs and infrastructure constraints around a proposed configuration.
Engineering validation, component sourcing and written project terms must precede any delivery commitment.
04 / A FEW DETAILS
Not yet. The collection shows exterior concepts and target configurations. Final specifications, pricing, availability, warranty and shipping terms have not been announced. This website does not accept orders or payments.
W1 targets one 300 W RTX PRO 6000 Blackwell Max-Q Workstation Edition GPU. W2 targets one 600 W Workstation Edition GPU. W3 targets two 600 W Workstation Edition GPUs. Each card has 96 GB of memory; these are GPU specifications, not whole-system power figures.
Each W3 GPU has separate 96 GB memory. Whether an application can distribute work or data across cards depends on the application, model and parallelization strategy. GPU count does not establish linear performance scaling.
S4 targets H200 NVL PCIe GPUs on a supported system platform. S8 targets a complete HGX H200 platform with SXM GPUs and its associated baseboard and interconnect. They are different platform formats and cannot be substituted as interchangeable add-in cards.
Share the applications you use, your workload, preferred model, installation country, project scale and timing. For servers, include available rack space, power and cooling where known. An inquiry starts a discussion and does not reserve hardware.
YOUR NEXT SYSTEM STARTS WITH THE WORK
Tell us about your applications, your project and the system you have in mind. Let's start with the requirements.