Whole-system understanding · Self-paced virtual instruction

AI Data-Center Campus: From Grid to GPU

Walk the complete campus: substations, backup power, data halls, cooling and shared services. Start the walkthrough below, then explore or test your understanding.

8 detailed lessons2 evidence-based case files60–85 minutes suggestedIncluded with All-Access
About this training · outcomes & learning route

Explore campus power, cooling and computing with a narrated system overview, eight close-up 3D lessons, a system diagram and two troubleshooting cases.

What you will practise

  • Trace the electrical path from utility interconnection to IT equipment and distinguish backup roles.
  • Follow heat from processors through rack liquid cooling, plant loops and campus heat rejection.
  • Identify shared dependencies across data halls, power, water, controls and communications.
  • Interpret capacity, demand, availability and measurement boundaries without equating them.
  • Investigate two campus-level evidence cases and prepare an actionable learning handover.

Starting point: Basic electrical-power and heat-transfer concepts. No server-administration experience is required. The campus is one fictional arrangement, not a representation of every AI deployment.

Use the 3D virtual instructor, orbit and zoom controls, full lesson transcripts, recorded Cedar narration in American English, knowledge checks and browser-local learning notes. Start narrated walkthrough to visit all eight subsystem steps automatically. The instructor travels to each teaching location before the explanation starts. Transits are illustrative and time-compressed; they do not validate real access or clearances.

Your learning route

  1. 1. Read the campus from the utility boundary

    Start with the whole campus. The substation connects utility supply to the site distribution arrangement, but the data halls depend on much more than electricity at the boundary. Cooling, controls, water, communications and access support the same service. Find the power yard, mechanical plant and buildings before zooming in on a single rack.

  2. 2. Follow the electrical path through switchgear

    Campus switchgear distributes and controls defined electrical paths. Locate the incoming and outgoing sections and compare the routes serving the halls and mechanical plant. A feeder serving a cooling pump can be just as consequential to computing service as a feeder serving IT equipment. Identify the intended dependency before treating circuits as interchangeable spare capacity.

  3. 3. Understand what backup generation actually supports

    Backup generators support defined loads through the campus distribution and transition arrangement. Inspect the engine enclosure, cooling/exhaust features and connected electrical equipment. Generator start, acceptable output and a load receiving that output are different states. A machine that is running in the yard has not necessarily restored every service in the hall.

  4. 4. Distinguish UPS support from long-duration backup

    The uninterruptible power supply supports its defined electrical load within the capability of the installed arrangement. Inspect the UPS equipment and associated battery area, then follow its output toward distribution and racks. It performs a different role from the generator. Neither the number of cabinets nor their visual size establishes actual support duration.

  5. 5. Connect the rack to power, cooling and data

    Enter the data-hall view and inspect the GPU-rack rows, overhead distribution and cooling interfaces. Each rack converts electrical input into computing work and heat. A rack is therefore part of at least three service paths: electrical supply, heat removal and data communication. A normal power indication does not establish that every workload is receiving the intended service.

  6. 6. Trace the liquid-cooling interface

    The coolant distribution unit connects defined cooling circuits serving IT equipment. Inspect its enclosure, pumps and paired supply/return connections, then trace toward the racks and the facility loop. Depending on design, heat exchange separates fluids with different quality and operating requirements. Similar-colored lines should not be assumed to contain interchangeable water.

  7. 7. Follow the cooling plant under the stated load

    The cooling plant supports the campus thermal demand through its actual arrangement of chillers, pumps, heat exchangers and controls. Locate the plant and trace its supply toward the data halls. Capacity is not the same as current demand, and installed equipment is not the same as available equipment. A unit can be physically present while unavailable for maintenance.

  8. 8. Finish at heat rejection and the shared campus services

    Heat must leave the campus cooling system or reach a useful heat-recovery destination. Inspect the heat-rejection equipment and trace it back toward the plant. Then locate the water-service area and network-operations building. These shared services help explain why a campus should be inspected as a connected whole rather than as independent buildings.

Troubleshooting practice

Case 1: rack temperatures rise in two halls

Two fictional halls report warmer rack cooling conditions during comparable workload. Use the supplied campus records to locate the most defensible investigation boundary.

Case 2: running generators, unrecovered mechanical feeder

A fictional utility event leaves IT service on UPS support while a mechanical feeder has not recovered as expected. Investigate the difference between generator availability and completed power delivery.

Review supplied evidence records, inspect the associated equipment, compare explanations and choose a responsible next action. These case files do not inject faults into the explorer or generate live simulated measurements.

Scope and learning references

This is a fictional campus teaching model, not a digital twin, certified availability design or engineering study. It does not solve power flow, fault current, protection coordination, transfer timing, battery autonomy, fluid networks, CFD, water treatment or chip temperatures. Drawn A/B paths and spare equipment do not establish failure independence or a redundancy rating. Case records are authored examples, not live simulated measurements.

This is scripted virtual instruction, not a live engineer, an AI conversation or a professional qualification. Completion records practice milestones only. Alex’s Cedar narration is AI-generated. All text remains readable without audio or 3D.

Reviewed November 19, 2026. Equipment-specific documents and applicable editions govern actual installations.

Continue learning

Related engineering studio · All instructor-led training