--- title: "How Modular AI Data Centers Are Built and Commissioned | Aguasto" description: "Follow the handoffs between requirements, factory integration, transport, site acceptance and compute-service evaluation without treating any single milestone as completion." canonical: "https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning" last-updated: "2026-10-13" --- # How Modular AI Data Centers Are Built and Commissioned | Aguasto Canonical HTML: [How Modular AI Data Centers Are Built and Commissioned](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning) > Follow the handoffs between requirements, factory integration, transport, site acceptance and compute-service evaluation without treating any single milestone as completion. [Aguasto](https://www.aguasto.com/) [learn](https://www.aguasto.com/learn) How Modular AI Data Centers Are Built and Commissioned [Resources](https://www.aguasto.com/resources) [GPU Library](https://www.aguasto.com/gpus) [Model Index](https://www.aguasto.com/models) [Inference Stack](https://www.aguasto.com/inference-stack) [Learn](https://www.aguasto.com/learn) [Insights](https://www.aguasto.com/insights) [Research](https://www.aguasto.com/research) [Engineering tools](https://www.aguasto.com/tools) Aguasto / learn ## How Modular AI Data Centers Are Built and Commissioned Follow the handoffs between requirements, factory integration, transport, site acceptance and compute-service evaluation without treating any single milestone as completion. Original Aguasto conceptual illustration. Not manufacturer geometry, an operating Aguasto asset, or a measured benchmark. By Aguasto Research · Published October 13, 2026 · Revision 1 Aguasto Research documents third-party hardware, models and software. References do not represent Aguasto inventory, evaluated deployments or available inference services. Illustration: How Modular AI Data Centers Are Built and Commissioned: original technical study. Requirements split into factory and site work. The branches meet at site integration, followed by integrated acceptance and a separate compute-service evaluation. Requirements split into factory and site work. The branches meet at site integration, followed by integrated acceptance and a separate compute-service evaluation. Original Aguasto illustration. Read the guide for configuration and measurement limitations. ### Start with the acceptance boundary Vertiv’s commissioning material separates preparation, checks and integrated assessment rather than treating delivery as a single event. Its prefabricated facility material describes an integrated package, but the package must still join a particular site. The practical implication is to define the intended acceptance evidence while requirements are being agreed, instead of leaving completion criteria until the end. Our suggested record names the proposed equipment configuration, the site interfaces, the parties reviewing each boundary and the evidence required at handover. Separate supplier factory acceptance, site equipment acceptance, integrated facility behavior and customer workload evaluation. These checkpoints answer different questions. An acceptance record should also identify unresolved items and the decision authority for closing them. A signed delivery receipt cannot answer all of those questions, even when equipment arrived undamaged and on schedule. [Commissioning your data center for greater availability and safety — Vertiv (external source)](https://www.vertiv.com/globalassets/services/services/project-services/commissioning-your-data-center-for-greater-availability-safety.pdf) · [Prefabricated Data Center brochure — Vertiv (external source)](https://www.vertiv.com/4a7618/globalassets/shared/vertiv-prefabricated-data-center-brochure.pdf) ### Factory integration has a bounded test scope Prefabrication can move selected assembly and testing into a factory environment. OCP’s liquid-cooling integration guidance discusses manufacturing and testing before transport as part of a broader logistics process. That evidence is useful because it records what the supplier assembled and checked. Its scope depends on the equipment, conditions and interfaces present during the test. A buyer should request a plain-language test boundary: what configuration was tested, what was simulated, what remained outside the factory and what exceptions were recorded? Carry those limitations into the site plan. Our analysis favors a chain of evidence in which each later checkpoint closes earlier uncertainties. It would be misleading to call a factory result a complete site result when the actual utility supply, heat rejection, network paths or installed configuration were absent. [Liquid Cooling Integration and Logistics White Paper, revision1.0 — Open Compute Project (external source)](https://www.opencompute.org/documents/ocp-liquid-cooling-integration-and-logistics-white-paper-revision-1-0-1-pdf) · [Network edge enabled by prefabricated modular data centers — Schneider Electric (external source)](https://blog.se.com/datacenter/2022/05/17/supporting-network-edge-prefabricated-modular-data-centers/) ### Treat transport as a handoff OCP’s logistics discussion covers handling and transport of liquid-cooled equipment and its later integration. These are not merely commercial shipping details. Preparation, receipt and reinstatement need coordination with the supplier’s instructions and the project’s qualified engineering team. This article does not provide transport, lifting or electrical work procedures. For planning, maintain a transfer record connecting the factory configuration to the equipment received on site. Record the responsible parties, inspection evidence and any changes or exceptions. A useful commercial question is who owns a discrepancy discovered after delivery, and how acceptance pauses until that discrepancy is assessed. Preserve the source documents rather than replacing them with a vague completion label. The objective is traceability across the handoff, not a promise that every module follows an identical transport process. [Liquid Cooling Integration and Logistics White Paper, revision1.0 — Open Compute Project (external source)](https://www.opencompute.org/documents/ocp-liquid-cooling-integration-and-logistics-white-paper-revision-1-0-1-pdf) ### The site join is an integration milestone The prefabricated package meets external civil, electrical, cooling and network interfaces at the installation site. Vertiv’s startup and commissioning services distinguish equipment startup from commissioning activities. A useful readiness plan therefore shows required external dependencies before the package can be evaluated under its intended operating conditions. Prefabrication does not remove the need for site approvals and qualified installation. Our interface register asks whether each upstream service matches the accepted design, whether the installed equipment matches the factory record and which open items prevent the next checkpoint. Keep responsibility visible across different suppliers. A completed enclosure can coexist with an incomplete external dependency. Conversely, an energized site can still lack accepted equipment. Progress reporting becomes more useful when it names the boundary that has been completed instead of describing the whole project as ready. [Prefabricated Data Center brochure — Vertiv (external source)](https://www.vertiv.com/4a7618/globalassets/shared/vertiv-prefabricated-data-center-brochure.pdf) · [Startup and Site Acceptance Test — Vertiv (external source)](https://www.vertiv.com/en-us/services-catalog/services/project-services/startup-and-commissioning-services/) ### Assess integrated behavior Vertiv’s commissioning guidance frames integrated testing as a way to evaluate facility behavior, including interactions among systems. Equipment-level startup does not by itself demonstrate the response of the entire installation. The project’s qualified commissioning team must establish the applicable test scope and safe procedures; a general website guide cannot prescribe them. For procurement and governance, ask which scenarios were assessed, what assumptions and temporary arrangements applied, which exceptions remain and how the operator received the results. A report should make pass criteria and configuration identity inspectable. Our proposed status model distinguishes checked equipment, assessed integrated behavior and accepted operating handover. This avoids turning an isolated successful test into a claim that all normal and abnormal operating conditions have been proven. [Commissioning your data center for greater availability and safety — Vertiv (external source)](https://www.vertiv.com/globalassets/services/services/project-services/commissioning-your-data-center-for-greater-availability-safety.pdf) ### Connect the facility to a compute product A facility handover and a customer compute acceptance are separate boundaries in our proposed operating model. A future dedicated cluster evaluation would need an agreed system configuration and workload criteria. A managed inference evaluation would additionally need a checkpoint, runtime, request distribution and quality boundary. Those decisions belong in a service acceptance record rather than being inferred from building power or rack count. Use the existing inference benchmark methodology to define measurements before any test runs. State what the buyer observes, which failures count, how concurrency is controlled and what constitutes a completed response. Keep a facility milestone and a software-service milestone in separate fields. Aguasto has not executed the commissioning or GPU workload evaluations described here. The publication provides a method for preparing evidence, not operating instructions or a report of completed Aguasto work. ### Scope & limitations - This is a planning and evidence guide, not a commissioning procedure or safety instruction. - Factory, site and compute-service acceptance have different boundaries. - No duration, certified design or executed Aguasto acceptance test is claimed. ### Continue the research [What Is a Modular AI Data Center?](https://www.aguasto.com/learn/what-is-a-modular-ai-data-center) [Why tokens per second is not enough to benchmark inference](https://www.aguasto.com/research/inference-benchmark-methodology) [Why Electricity Availability Shapes AI Infrastructure Deployment](https://www.aguasto.com/research/electricity-availability-and-time-to-power) [The Aguasto modular thesis](https://www.aguasto.com/why-modular) [Potential OEM and strategic collaboration](https://www.aguasto.com/partners) ### Sources & verification The register preserves each source's publication and verification date. Manufacturer claims and documentation are attributed; Aguasto has not run the cited products in production. - [Commissioning your data center for greater availability and safety — Vertiv ↗](https://www.vertiv.com/globalassets/services/services/project-services/commissioning-your-data-center-for-greater-availability-safety.pdf) Commissioning scope spans requirements/design/construction/operations; Vendor startup, integrated test and turnover are distinct. Publication: date not stated in the reviewed source. Verified: 2026-10-13. - [Prefabricated Data Center brochure — Vertiv ↗](https://www.vertiv.com/4a7618/globalassets/shared/vertiv-prefabricated-data-center-brochure.pdf) Factory/site/service integration phases; Engineering, acceptance plans, installation and commissioning roles. Publication: date not stated in the reviewed source. Verified: 2026-10-13. - [Liquid Cooling Integration and Logistics White Paper, revision1.0 — Open Compute Project ↗](https://www.opencompute.org/documents/ocp-liquid-cooling-integration-and-logistics-white-paper-revision-1-0-1-pdf) FWS/TCS/CDU terminology and liquid-to-liquid isolation; Manufacturing, transport, inspection and commissioning handoffs; Wetted materials and project-specific tests. Publication: date not stated in the reviewed source. Verified: 2026-10-13. - [Network edge enabled by prefabricated modular data centers — Schneider Electric ↗](https://blog.se.com/datacenter/2022/05/17/supporting-network-edge-prefabricated-modular-data-centers/) Factory assembly and site work may proceed in parallel. Publication: 2022-05-17. Verified: 2026-10-13. - [Startup and Site Acceptance Test — Vertiv ↗](https://www.vertiv.com/en-us/services-catalog/services/project-services/startup-and-commissioning-services/) Electrical/mechanical startup checks; Functional testing in operational modes and integrated site acceptance. Publication: date not stated in the reviewed source. Verified: 2026-10-13. ### In this publication [Start with the acceptance boundary](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning#section-1) [Factory integration has a bounded test scope](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning#section-2) [Treat transport as a handoff](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning#section-3) [The site join is an integration milestone](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning#section-4) [Assess integrated behavior](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning#section-5) [Connect the facility to a compute product](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning#section-6) [More learn ↗](https://www.aguasto.com/learn) [Source register ↓](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning#sources) ## Company navigation - [Home](https://www.aguasto.com/) - [Infrastructure](https://www.aguasto.com/infrastructure) - [GPU Cloud](https://www.aguasto.com/gpu-cloud) - [AI Inference](https://www.aguasto.com/ai-inference) - [Development & Investment](https://www.aguasto.com/development) - [Company](https://www.aguasto.com/company) - [Contact](https://www.aguasto.com/contact) - [Privacy](https://www.aguasto.com/privacy) - [Resources](https://www.aguasto.com/resources) - [GPU Library](https://www.aguasto.com/gpus) - [Model Index](https://www.aguasto.com/models) - [Inference Stack](https://www.aguasto.com/inference-stack) - [Learn](https://www.aguasto.com/learn) - [Insights](https://www.aguasto.com/insights) - [Research](https://www.aguasto.com/research) - [Industry cases](https://www.aguasto.com/research/cases) - [Why Modular](https://www.aguasto.com/why-modular) - [Partners](https://www.aguasto.com/partners) - [GPU comparison](https://www.aguasto.com/gpus/compare) - [Model comparison](https://www.aguasto.com/models/compare) - [Engineering tools](https://www.aguasto.com/tools) - [Development friction](https://www.aguasto.com/research/hyperscale-development-friction) - [NVIDIA H100 SXM](https://www.aguasto.com/gpus/h100) - [NVIDIA H200 SXM](https://www.aguasto.com/gpus/h200) - [NVIDIA B200 SXM](https://www.aguasto.com/gpus/b200) - [NVIDIA B300](https://www.aguasto.com/gpus/b300) - [NVIDIA GB200 NVL72](https://www.aguasto.com/gpus/gb200-nvl72) - [NVIDIA GB300 NVL72](https://www.aguasto.com/gpus/gb300-nvl72) - [NVIDIA Vera Rubin NVL72](https://www.aguasto.com/gpus/vera-rubin-nvl72) - [AMD Instinct MI300X](https://www.aguasto.com/gpus/mi300x) - [AMD Instinct MI325X](https://www.aguasto.com/gpus/mi325x) - [AMD Instinct MI355X](https://www.aguasto.com/gpus/mi355x) - [AMD Instinct MI455X](https://www.aguasto.com/gpus/mi455x) - [AMD Helios](https://www.aguasto.com/gpus/helios) - [GLM-4.5](https://www.aguasto.com/models/glm-4-5) - [Qwen3-235B-A22B](https://www.aguasto.com/models/qwen3-235b-a22b) - [DeepSeek-R1](https://www.aguasto.com/models/deepseek-r1) - [Llama 3.3 70B Instruct](https://www.aguasto.com/models/llama-3-3-70b-instruct) - [Gemma 3 27B IT](https://www.aguasto.com/models/gemma-3-27b-it) - [gpt-oss-120b](https://www.aguasto.com/models/gpt-oss-120b) - [Kimi-K2-Instruct](https://www.aguasto.com/models/kimi-k2-instruct) - [GLM-5.3](https://www.aguasto.com/models/glm-5-3) - [Qwen3.8-27B](https://www.aguasto.com/models/qwen3-8-27b) - [DeepSeek-V4.1-Flash](https://www.aguasto.com/models/deepseek-v4-1-flash) - [Llama 4 Scout 17B-16E Instruct](https://www.aguasto.com/models/llama-4-scout-17b-16e-instruct) - [Gemma 4 31B IT](https://www.aguasto.com/models/gemma-4-31b-it) - [Kimi K3](https://www.aguasto.com/models/kimi-k3) - [vLLM](https://www.aguasto.com/inference-stack/vllm) - [SGLang](https://www.aguasto.com/inference-stack/sglang) - [TensorRT-LLM](https://www.aguasto.com/inference-stack/tensorrt-llm) - [NVIDIA Dynamo](https://www.aguasto.com/inference-stack/dynamo) - [NVIDIA Triton Inference Server](https://www.aguasto.com/inference-stack/triton) - [AMD ROCm](https://www.aguasto.com/inference-stack/rocm) - [Ray Serve](https://www.aguasto.com/inference-stack/ray-serve) - [Kubernetes GPU orchestration — NVIDIA device plugin](https://www.aguasto.com/inference-stack/kubernetes) - [Model weights are only the beginning of the memory budget](https://www.aguasto.com/learn/model-weights-and-runtime-memory) - [Why tokens per second is not enough to benchmark inference](https://www.aguasto.com/research/inference-benchmark-methodology) - [Prefill, decode and the cost of separating them](https://www.aguasto.com/learn/prefill-decode-and-disaggregated-serving) - [Choosing between vLLM, SGLang and TensorRT-LLM](https://www.aguasto.com/learn/choosing-an-inference-serving-framework) - [From Power to Inference—Building a Distributed AI Compute Network](https://www.aguasto.com/learn/from-power-to-inference) - [Dedicated compute and shared inference: compare the whole workload](https://www.aguasto.com/insights/dedicated-and-shared-inference-economics) - [Distributed inference reliability begins with the failure boundary](https://www.aguasto.com/learn/distributed-inference-reliability) - [Core Scientific: reading contracted, energized and billed capacity](https://www.aguasto.com/research/cases/core-scientific-capacity-disclosures) - [Why Modular AI Infrastructure Matters Now](https://www.aguasto.com/research/why-modular-ai-infrastructure-matters-now) - [What Is a Modular AI Data Center?](https://www.aguasto.com/learn/what-is-a-modular-ai-data-center) - [How Modular AI Data Centers Are Built and Commissioned](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning) - [Liquid Cooling for High-Density GPU Systems: Follow the Interfaces](https://www.aguasto.com/learn/liquid-cooling-gpu-rack-interfaces) - [Why Electricity Availability Shapes AI Infrastructure Deployment](https://www.aguasto.com/research/electricity-availability-and-time-to-power) - [What Is KV Cache? Capacity, Reuse and Isolation](https://www.aguasto.com/learn/kv-cache-capacity-and-policy) - [Continuous Batching: Throughput Without Hiding Latency](https://www.aguasto.com/learn/continuous-batching-and-latency) - [Applied Digital: Reading the Polaris Forge 1 Delivery Milestones](https://www.aguasto.com/research/cases/applied-digital-polaris-forge-capacity-milestones) - [Crusoe Abilene: Separate Delivered Infrastructure from Expansion Plans](https://www.aguasto.com/research/cases/crusoe-abilene-delivery-and-expansion) - [Official LinkedIn](https://www.linkedin.com/company/aguasto) This is a generated representation of the public HTML page. Diagrams are conceptual. GPU and inference services are planned and are not currently available to book or call. The contact page describes the available inquiry flow.