Question 1

What is Leviathan Systems?

Accepted Answer

Leviathan Systems is a GPU infrastructure deployment company. We assemble, cable, test, and commission GPU clusters for AI-scale data centers across the United States. Our scope covers the full physical build — from bare rack to production-ready cluster — across every current NVIDIA platform from H100 through GB300 NVL72. We are a pure-services company. We do not sell or resell hardware. Clients procure their own equipment from OEMs like Dell, NVIDIA, Supermicro, and Arista, and we handle the physical deployment.

Question 2

What services does Leviathan Systems provide?

Accepted Answer

We provide four core infrastructure services: GPU rack assembly and integration (rail install, server and switch placement, power and network cabling, NVLink and InfiniBand routing, POST verification), structured cabling and fiber (OM4/OM5 multimode, OS2 single-mode, MPO/MTP trunking, DAC, AOC, and AEC interconnects to TIA-942 and BICSI standards), network testing and commissioning (OTDR testing, insertion loss and return loss verification, copper certification, full documentation with test reports and cable maps), and liquid cooling integration (CDU installation, rack-level manifold routing, quick-disconnect fittings, leak detection, and facility water loop coordination).

Question 3

Where does Leviathan Systems operate?

Accepted Answer

We deploy GPU infrastructure nationwide across the United States. We mobilize crews to any US data center market within one week of contract execution.

Question 4

How large is the Leviathan Systems team?

Accepted Answer

We maintain 30 to 100 field technicians depending on active project load. Our team members have hands-on deployment experience at facilities operated by Meta, Oracle, xAI, and Computacenter. We are operator-led — our founders are on-site during deployments, not managing remotely.

Question 5

What is Leviathan's track record?

Accepted Answer

We have assembled over 1,500 GPU racks, deployed more than 25,000 cable connections, and completed 8+ data center projects. We have deployed infrastructure across every current NVIDIA platform generation — H100, H200, GH200, GB200 NVL72, and GB300 NVL72. We are currently executing active GB300 NVL72 deployments at hyperscale AI training facilities.

Question 6

What does GPU rack assembly include?

Accepted Answer

Our rack assembly process covers the full mechanical and electrical build of a GPU rack: rail installation, server and switch placement, power cable routing, network cable routing, NVLink interconnect installation, InfiniBand cabling, cable management and labeling, and POST verification. We handle DGX, HGX, GB200 NVL72, and GB300 NVL72 platforms from bare rack to power-on.

Question 7

Which NVIDIA platforms do you deploy?

Accepted Answer

We deploy all current NVIDIA platforms: H100 (Hopper, 80GB HBM3, 700W TDP, air or liquid cooled), H200 (Hopper with HBM3e, 141GB memory, drop-in H100 upgrade), GH200 (Grace Hopper Superchip, unified CPU+GPU, NVLink-C2C), GB200 NVL72 (72 Blackwell GPUs per rack, 120kW, 100% liquid cooled), and GB300 NVL72 (Blackwell Ultra, 288GB HBM3e per GPU, currently deploying). Our teams are trained on the specific assembly procedures, cabling topologies, and commissioning requirements for each platform.

Question 8

Which OEM hardware do you work with?

Accepted Answer

We deploy infrastructure built on hardware from Dell, NVIDIA, Supermicro, and Arista. This includes NVIDIA DGX and HGX systems, Supermicro GPU server platforms, and Arista switching infrastructure. We are hardware-agnostic on the services side — if it racks, cables, and powers on, we deploy it.

Question 9

What is POST verification?

Accepted Answer

POST (Power-On Self-Test) verification is the final step in our rack assembly process. After all physical installation and cabling is complete, we power on every server, validate that all GPUs are detected, confirm NVLink and InfiniBand connectivity, verify firmware versions, and document the results. No rack leaves our hands without confirmed POST.

Question 10

How many racks can you deploy per week?

Accepted Answer

Throughput depends on platform complexity and crew size. GB200 NVL72 and GB300 NVL72 racks require significantly more time than earlier platforms due to liquid cooling integration, higher cable density, and NVLink topology complexity. We scope timelines per project during the discovery phase.

Question 11

What cabling standards do you follow?

Accepted Answer

All installations conform to TIA-942 (Telecommunications Infrastructure Standard for Data Centers) and BICSI (Building Industry Consulting Service International) standards. These standards govern cable pathway design, bend radius, separation from power, labeling, and testing requirements.

Question 12

What types of fiber do you install?

Accepted Answer

We install OM4 and OM5 multimode fiber for short-distance GPU-to-switch connections, OS2 single-mode fiber for backbone and long-distance runs, and MPO/MTP trunk cables for high-density GPU cluster interconnects. We also deploy DAC (Direct Attach Copper), AOC (Active Optical Cable), and AEC (Active Electrical Cable) interconnects based on the switching topology and distance requirements.

Question 13

What is MPO/MTP trunking and why does it matter for GPU clusters?

Accepted Answer

MPO/MTP (Multi-fiber Push On/Multi-fiber Termination Push-on) connectors carry 8, 12, or 24 fibers in a single connector. In GPU cluster deployments, the sheer number of connections between GPU nodes and leaf/spine switches makes individual fiber runs impractical. MPO/MTP trunking consolidates these connections into high-density trunk cables that are pre-terminated, factory-tested, and faster to deploy. This reduces installation time, minimizes connection points (each of which is a potential failure), and produces cleaner cable management.

Question 14

Do you handle cable labeling and documentation?

Accepted Answer

Yes. Every cable we install is labeled at both ends with a standardized naming convention that maps to the network topology. We deliver complete cable maps and as-built documentation at project handoff. This documentation includes port-to-port mapping, cable types and lengths, test results, and rack elevation drawings.

Question 15

What testing do you perform on every connection?

Accepted Answer

Every fiber and copper connection is tested before handoff. Fiber testing includes OTDR (Optical Time Domain Reflectometer) testing to characterize the full link, insertion loss measurement, and return loss measurement. Copper testing includes certification to the appropriate category standard. Results are documented per connection and delivered as part of the project handoff package.

Question 16

What is OTDR testing?

Accepted Answer

OTDR (Optical Time Domain Reflectometer) testing sends a pulse of light down a fiber and measures the reflections that return. This produces a trace that shows the entire fiber link — every connector, splice, bend, and termination point — with distance and loss measurements at each event. OTDR testing identifies problems that simple pass/fail testing misses: a connector with acceptable insertion loss but a cracked ferrule, a fiber with a stress point that will fail under temperature change, or a splice with marginal performance.

Question 17

What is the difference between insertion loss and return loss?

Accepted Answer

Insertion loss measures how much optical power is lost as light passes through a connection point — lower is better. Return loss measures how much light is reflected back toward the source at a connection point — higher is better (meaning less reflection). Both measurements are critical for high-speed GPU interconnects. Excessive insertion loss reduces signal strength. Excessive return loss (low values) causes signal interference. We test both on every connection.

Question 18

Do you provide test reports and documentation?

Accepted Answer

Yes. Every project is delivered with a complete documentation package that includes individual test results for every fiber and copper connection, OTDR traces, cable maps showing port-to-port connectivity, rack elevation drawings, photographs of completed work, and a summary report. This documentation package is the standard deliverable that clients use for their own commissioning validation and ongoing operations.

Question 19

Why do GB200 and GB300 NVL72 racks require liquid cooling?

Accepted Answer

The GB200 NVL72 rack draws approximately 120kW of power. The GB300 NVL72 draws even more. At these power densities, air cooling cannot remove heat fast enough to keep GPUs within operating temperature. Direct liquid cooling circulates coolant through cold plates mounted directly on GPU and CPU packages, removing heat at the source with far greater efficiency than air. Every GB200 and GB300 deployment requires liquid cooling — there is no air-cooled option.

Question 20

What does liquid cooling integration include?

Accepted Answer

Our liquid cooling scope covers CDU (Coolant Distribution Unit) installation and connection to the facility water loop, rack-level manifold routing, quick-disconnect fitting installation at each server tray, leak detection system installation and validation, pressure testing and flushing of the cooling loop, coolant fill and flow balancing, and thermal performance validation under load. We coordinate with facility mechanical teams to ensure the building water supply meets the flow rate and temperature requirements of the GPU racks.

Question 21

What is a CDU and how does it work?

Accepted Answer

A CDU (Coolant Distribution Unit) is the interface between the facility chilled water loop and the rack-level cooling circuit. It contains a heat exchanger, pumps, filters, and controls. Facility water (typically 10-20°C) flows through one side of the heat exchanger. Rack coolant flows through the other side. The CDU regulates coolant temperature, flow rate, and pressure to keep GPUs within their thermal envelope. One CDU typically serves one or two GPU racks depending on the platform and thermal load.

Question 22

How do you handle leak detection in liquid-cooled deployments?

Accepted Answer

Every liquid-cooled deployment includes leak detection at multiple points: under each rack, at CDU connections, at manifold junctions, and at quick-disconnect fittings. Leak detection sensors are wired to monitoring systems that trigger alerts before any coolant reaches electrical components. We perform pressure testing on the cooling loop before introducing coolant, and we validate leak detection functionality as part of the commissioning process.

Question 23

How do I get started with Leviathan Systems?

Accepted Answer

Contact us through our website or email. We respond same day. The process from there is: discovery call to understand your project scope, platform, timeline, and facility requirements; site assessment if needed; detailed scope document with crew plan and timeline; contract execution; crew mobilization within one week. We do not require long procurement cycles. Once the scope is agreed, we move fast.

Question 24

How quickly can you mobilize a crew?

Accepted Answer

Under one week from contract execution to boots on the ground. For urgent deployments, we can mobilize faster. Our teams are structured for rapid deployment — we do not rely on third-party staffing agencies or subcontractor sourcing. When you engage Leviathan, you get our crew.

Question 25

What does a typical engagement look like?

Accepted Answer

A typical engagement follows five phases: Phase 1 is scoping — we review your hardware BOM, facility layout, and timeline. Phase 2 is planning — we produce a detailed work plan with crew allocation, sequencing, and milestones. Phase 3 is deployment — our crews execute rack assembly, cabling, and cooling integration. Phase 4 is testing — every connection is tested and documented. Phase 5 is commissioning and handoff — we deliver the documentation package, walk through the build with your operations team, and confirm production readiness.

Question 26

Do you work as a subcontractor for general contractors?

Accepted Answer

Yes. We regularly work as a specialty subcontractor for general contractors managing large data center builds. Our scope is the GPU-specific infrastructure: rack assembly, GPU cabling, NVLink/InfiniBand routing, liquid cooling integration, and testing. We integrate with the GC's project schedule and site protocols.

Question 27

What is your response time for inquiries?

Accepted Answer

Same day. We respond to every inquiry within business hours on the day it is received. If you submit a request in the evening, you will hear from us by the next morning.

Question 28

Do you provide ongoing support after deployment?

Accepted Answer

Our primary scope is deployment — assembly, cabling, testing, and commissioning. We deliver infrastructure that is production-ready at handoff. For clients requiring ongoing support, maintenance, or expansion work, we offer follow-on engagements scoped separately.

Question 29

How is Leviathan different from large staffing-model deployment companies?

Accepted Answer

Most large GPU deployment companies operate as staffing networks — they recruit technicians from a database and dispatch them to your site. Leviathan is operator-led. Our founders are on-site during deployments. Our crews are trained specifically on NVIDIA GPU platforms, not general IT infrastructure. We run smaller, tighter teams with direct quality control on every connection. You get a specialized GPU infrastructure crew, not a rotation of general-purpose technicians.

Question 30

Why does NVIDIA platform specialization matter?

Accepted Answer

Every NVIDIA platform has different assembly procedures, cabling topologies, power requirements, and cooling configurations. An H100 HGX rack has different NVLink routing than a GB200 NVL72. A GB300 NVL72 has different liquid cooling manifold architecture than a GB200. The jump from one generation to the next is not incremental — it is a fundamentally different build. Teams that deploy NVIDIA platforms as a specialty get these details right. Teams that treat GPU racks as 'just another rack and stack' introduce errors that surface during commissioning or, worse, during production.

Question 31

What OEM platforms has your team deployed?

Accepted Answer

Our team members have deployment experience across Meta, Oracle, xAI, and Computacenter facilities. We work with hardware from Dell, NVIDIA, Supermicro, and Arista. This cross-platform experience means we understand not just how to build a single rack, but how different vendor configurations interact within a cluster topology.

Question 32

Do you sell or resell GPU hardware?

Accepted Answer

No. Leviathan Systems is a pure-services company. We do not sell, resell, or broker GPU hardware, servers, switches, or any other data center equipment. Clients procure their own hardware through their OEM relationships. We deploy it. This keeps our incentives aligned — we have no reason to push specific hardware configurations. Our only job is to build it right.

Question 33

What does 'production-ready at handoff' mean?

Accepted Answer

It means that when we hand the infrastructure back to your operations team, every rack is assembled, every cable is installed and tested, every cooling loop is pressurized and validated, every server has passed POST verification, and the complete documentation package is delivered. No rework. No punch lists. Production-ready.

Question 34

Can you handle multi-platform deployments within the same facility?

Accepted Answer

Yes. Many data center builds involve multiple NVIDIA platforms deployed simultaneously or sequentially. A facility might have existing H100 infrastructure expanding with GB200 NVL72 racks, or a new build with both air-cooled H200 and liquid-cooled GB300 NVL72. We handle mixed-platform deployments with dedicated crews and work plans for each platform type.

Question 35

How does your quality control process work?

Accepted Answer

Quality control is embedded in every phase — assembly, cabling, testing, and commissioning. Every connection is inspected and verified before testing. Every fiber and copper link is individually tested with results recorded per connection. During commissioning, we perform a full system-level validation including POST on every server. Issues are resolved before handoff, not after. Our founders review project quality directly — there is no layer of middle management between the work and the people accountable for it.

Frequently Asked Questions_

Company & Overview

GPU Rack Assembly & Integration

Structured Cabling & Fiber

Network Testing & Commissioning

Liquid Cooling

Engagement & Process

Technical Differentiators

Still have questions? Talk to an engineer_