AI startups & inference providers
Discuss dedicated GPU infrastructure for model development, training and inference. Specify workload, memory needs and intended utilization.
Indicative build time: 3–8 weeks, plus shipping. Final schedule confirmed with your quote.
Planning a containerized data center for AI training, inference or GPU hosting? Request a 20 ft or 40 ft configuration. We review your company information, technical requirements and US delivery location before calculating equipment, shipping and related delivery costs. No fixed US shipping price is advertised.
For US companies evaluating containerized AI infrastructure—not consumer PCs or individual graphics cards. Tell us your workloads, GPU and rack requirements, available power, cooling needs and project timeline. We assess the configuration before preparing a project-specific quote. Globai is based in the Netherlands; US delivery is assessed per project.
Visit globanetweb.comDiscuss dedicated GPU infrastructure for model development, training and inference. Specify workload, memory needs and intended utilization.
Explore modular infrastructure for GPU hosting. Request a configuration based on your rack, network, cooling and power requirements.
Evaluate a containerized data center for your site. Share procurement requirements, available utilities and any required certifications for review.
Plan a 20 ft or 40 ft project with defined scope, delivery access and installation responsibilities. Capacity is engineered per project, not inferred from container length.
System in a system. Software creates demand for compute. Our own compute makes the software stronger. That is the loop we want to build.
The first compute step is deliberately simple: a planned modular container with 3 servers of 8 NVIDIA H200 GPUs each, physically located in the Netherlands. The capacity is intended for rental to AI teams and researchers, including via marketplaces such as Vast.ai.
The container is in planning and not yet operational. Delivery depends on financing, supply and permits.
Overview based on a supplied quotation for cooling, power and racks. Quantities are actual units; this is a planning specification, not an installation.
| # | Component | Qty | Specification |
|---|---|---|---|
| 01 | 42U server cabinet | Qty: 1× | 600 × 1200 × 2000 mm, load rating 2000 kg. Incl. 2 vertical cable channels, 4 PDU mounting plates, 1 side sealing kit, 1 cabinet connection kit, 50 M6 nuts. |
| 02 | Intelligent PDU | Qty: 2× | 63 A three-phase, 230/400 V, 20 × C13 + 4 × C19, RS485, main meter and main breaker. Sized for the initial approx. 30 kW, not full plant power. |
| 03 | Rack-mounted CDU | Qty: 1× | 130 kW, 442 × 810 × 264 mm. Deionised water, ethylene glycol or propylene glycol. |
| 04 | Manifold | Qty: 2× | 10 branches, UQD04 male, EPDM hose OD 45 mm, main pipe 40 × 40 mm stainless 304, nominal 15 L/min per branch. |
| 05 | Flexible hose | Qty: 20× | UQD04 female quick connector. Quantity tracks the branches on expansion. |
| 06 | Dry cooler | Qty: 1× | 164.3 kW, fluid 46/35 °C, rated flow 13.9 m³/h. Rated at design conditions, not guaranteed for every ambient temperature. |
| 07 | Modular UPS frame | Qty: 1× | 150 kVA frame, 3-phase in/out, ±240 VDC (30–50 batteries), crest factor 3:1, RS485/RS232/dry contact/SNMP/EPO. 600 × 1100 × 1600 mm, 178 kg. Cabinet reserved for approx. 100 kW expansion. |
| 08 | UPS module 25 kVA/25 kW | Qty: 3× | 75 kW in the specified configuration (planned site, not installed); 3 more modules possible. With N+1 across 3 modules, effectively approx. 50 kW. |
| 09 | Lithium battery system | Qty: 1× | 563.2 V, 100 Ah. Supplier states >30 min at 50 kW, >15 min at 100 kW. Battery estimate, not a UPS output approval. |
| 10 | Monitoring | Qty: 1× | Temperature/humidity, leak detection, remote via SNMP. Estimated configuration. |
Additional cabinets are possible; the exact rack count is not yet fixed.
These components can be purchased, rented or leased as additions. A credit check is required.
Servers and GPUs can be purchased, rented or leased as additions. A credit check is required. Planned configuration: 3 × 8 NVIDIA H200 in the Netherlands, with compute capacity for rental via Vast.ai.
The quotation mentions 20 ft and 40 ft only as sea-freight shipping formats, not as a turnkey container bill of materials. A maximum cannot be inferred from the length alone, and 40 ft does not automatically double capacity.
Sent along with the form. Indicative, not a binding order.
For the container infrastructure, Globai is evaluating a specialised manufacturer of modular data centre solutions. The photos below are supplier reference images. They show the intended construction method, not a Globai installation.
The ambition is step by step. Each next step is taken only once the previous one is operationally proven and financing is in place.
Active services for clients. A source of knowledge and relationships.
24 H200 GPUs in the Netherlands, intended for rental to AI teams.
Expansion after proven utilisation and operations.
Multiple containers, linked to our own AI software.
Globai is led by its founder, who serves clients, builds software and developed the compute plans himself. Short lines: whoever talks to Globai talks directly to the decision-maker.
2 concept visualisations and 7 reference photos from the supplier. Click to enlarge; navigate with the arrow keys.
Yes. US businesses can request a 20 ft or 40 ft container configuration. Share your legal company name, business address, contact details, delivery location and technical requirements. US shipping and related costs are calculated after reviewing the complete company and project information.
There is no fixed shipping price. Freight, insurance, customs handling, applicable duties or taxes, inland delivery and any unloading or installation scope must be assessed for your destination and agreed delivery terms. The quotation will identify the included scope and responsibilities; submitting an inquiry does not book shipping.
Provide your legal company name, registered business address, website and registration reference if available, contact person and business email, plus the delivery country, state, city, ZIP code and site address. Include container size, power supply, cooling needs, GPU requirements, access or unloading constraints and target date. Do not send tax documents, credit reports, bank details or other sensitive records through this form.
Capacity depends on server dimensions, racks, power, cooling and service access. A 40 ft container does not automatically double capacity. The illustrative 24 × NVIDIA H200 project is planned in the Netherlands, not a guaranteed specification for every US order.
Servers, GPUs and additional components can be requested for purchase, rental or lease. A credit check is required. Availability, eligibility and final terms are reviewed for each project; submitting a request does not guarantee financing approval.
No US certification, permitting or site compatibility is guaranteed by this website. Tell us the electrical, safety and local approval requirements for your location so they can be reviewed. The final quote defines equipment, delivery and installation responsibilities.
No. The illustrated modular container with 24 NVIDIA H200 GPUs is a project in planning. It is not an operational US data center or ready-to-ship inventory.
Via the contact form. Timeline, structure and terms are only discussed in person. This website is not an offer of securities and promises no returns.
Choose 20 ft or 40 ft and share your company and delivery information. Equipment, US shipping and related delivery costs are calculated after reviewing the complete details. This is a quote request, not a purchase or shipping booking.
This page is not an offer of securities and contains no promise of returns.
On the home page only, your browser asks GeoJS (get.geojs.io) for your country code based on your IP address to suggest a language. We do not store IP addresses; only your language choice is saved locally in your browser.