Power density is the amount of electrical load assigned to a rack or a room. It matters because almost all of that electricity becomes heat. A building designed for 10 kW racks can be operationally overwhelmed by a small number of 80 kW or 100 kW AI racks even if the campus average still looks moderate.

The 2024-to-2027 period is important because AI systems are shifting from individual accelerator cards toward integrated rack-scale systems. The rack is becoming a unit of deployment, power distribution, cooling, networking, and service.

The density curve is not uniform

A conventional enterprise rack may operate below 10 kW. A virtualized cloud rack may run in the 10-to-20 kW range. High-performance computing and AI racks can move through 30, 50, 80, and above 100 kW depending on the server design and workload.

Vertiv’s 2025 analysis cited a change in the installed mix: below-10-kW racks represented 38% of capacity, down from 47% in 2024; 10-to-20-kW racks represented 30%, and 20-to-30-kW racks represented 28%. The exact mix varies by provider and building, but the direction shows why legacy air designs are being reworked.

A high-density rack is not always full

Nameplate power is a ceiling, not necessarily the continuous draw. A rack may be provisioned for 80 kW and average 55 kW. It may also ramp quickly as workloads start. The electrical design must address both sustained and transient demand.

A facility that uses average power for planning can underbuild distribution, generators, cooling, or UPS systems. A facility that assumes every rack will draw its maximum can overbuild expensive infrastructure. The solution is measurement and a clear operating envelope: minimum, typical, peak, ramp rate, and allowable diversity.

What changed in rack-scale AI

NVIDIA’s reference material for NVL72 systems describes a rack containing 72 GPUs, 18 compute trays, nine NVLink switch trays, and management switches. A later Vera Rubin NVL72 user guide lists four power shelves, each rated at 110 kW, for a total shelf rating of 440 kW before considering actual load, redundancy, and facility losses.

That is not a claim that every rack continuously consumes 440 kW. It illustrates a new design scale. A single rack can now be a large electrical load with liquid manifolds, a bus bar, high-speed networking, and maintenance procedures that look more like a small modular system than a traditional server cabinet.

A facility must plan the electrical path

Power density becomes a sequence of limits: utility service, substation, medium-voltage distribution, transformer, switchgear, remote power panels, busway, rack power shelves, and the server power supplies. The weakest link sets the usable density.

Redundancy complicates the picture. A rack may have two power paths, but that does not mean the site can use twice the connected capacity. The operator needs to know the failure case: can one path carry the rack, can generators pick up the load, and how much spare capacity remains after maintenance?

Power conversion also produces heat. A rack’s IT load is not the same as the utility draw. UPS losses, transformers, power shelves, pumps, and networking equipment add overhead. PUE can improve with better design, but the facility still has to size conductors and cooling for the real worst case.

Cooling sets the practical limit

Air cooling can work at moderate density with good containment and high supply airflow. At higher density, fans and air movers consume more energy, and the room becomes sensitive to layout and bypass airflow.

Liquid cooling moves the heat with a smaller physical path. Direct-to-chip cooling may handle the hottest components, while rear-door heat exchangers remove residual heat. The facility then needs coolant distribution units, pumps, heat exchangers, leak detection, and service isolation.

The best design is often mixed. A 2027 hall may contain low-density storage racks, 30 kW CPU racks, and 100 kW AI racks. The design should not force every rack into the most expensive cooling architecture, but it must prevent low-density assumptions from controlling the high-density row.

A reasonable planning framework for 2024-2027

For 2024, treat 10-to-30 kW racks as a mainstream planning range and higher-density AI deployments as specialized zones. For 2025, assume more 30-to-80 kW racks and require liquid-ready distribution in new AI rooms. For 2026 and 2027, plan rack-scale systems explicitly, including 100-kW-class electrical and cooling branches where the selected hardware requires them.

Those are planning bands, not universal forecasts. Hardware release schedules differ, and some systems are more efficient than others. The underlying point is that a site should not promise a density it cannot prove through a full-load test.

The commercial implication

Power density affects the value of a building. A site with 100 MW of utility service is not equivalent to a site that can safely deliver 100 MW to high-density AI racks. The second site has the distribution, cooling, commissioning, controls, and maintenance capability to turn connection into usable compute.

Investors and tenants should ask for rack-level test data, not just a campus megawatt number. The important evidence is energization, cooling performance, redundancy, and the number of racks that can run simultaneously.

A site plan should also include a density migration path. The first room may begin with 30-kW racks, then add 60-kW racks after the coolant loop and switchgear are upgraded. That sequence is more credible than claiming the whole building is AI-ready on day one. Commissioning should prove the highest-density row, the failure of one power path, the loss of one cooling component, and the restart of a synchronized workload. The result is a usable capacity number rather than a brochure number.

A data center is not a single asset with a single lifespan. It is a set of layers that wear out at different speeds. The concrete shell may serve for decades, the electrical and cooling systems for a shorter period and the servers inside for only a few years. Understanding those different clocks explains a lot about ongoing capital spending in the industry.

Three clocks running at once

Each layer of a data center follows its own timeline:

• The building and site, including foundations, structure and roof, can last for decades with maintenance
• Electrical and mechanical systems, such as generators, switchgear, chillers and cooling units, typically need major refurbishment or replacement on a shorter cycle
• Servers, storage, networking and accelerators are refreshed most often, because newer equipment offers better performance and efficiency

These timelines also explain why accounting depreciation schedules differ so widely across asset categories.

Refresh cycles

When servers are replaced, the old equipment is removed and new equipment is installed, often without major changes to the building. This refresh cycle is one of the main reasons a data center’s total spending over its life can far exceed its original construction cost.

Retired servers may be resold, redeployed for less demanding work or recycled. Data storage devices are typically wiped or physically destroyed to protect information, following established data sanitization practices.

Retrofits for new technology

Sometimes the building outlives the design it was built for. The rise of AI hardware, with much higher power density per rack, has pushed operators to consider retrofitting older facilities. Retrofits might include adding liquid cooling, upgrading electrical distribution, reinforcing floors or reconfiguring halls.

Retrofitting can be cheaper and faster than building new, especially where a site already has a valuable utility connection. But there are limits: an older building may lack the structural capacity, ceiling height or power to support the newest equipment.

Decommissioning

Eventually, a facility may reach the end of its useful life for its original purpose. Decommissioning involves removing equipment, safely handling materials such as batteries, fuels and refrigerants under applicable environmental rules, and either repurposing or demolishing the building. Some older data centers have been converted to other uses, and their sites may be redeveloped with new facilities that reuse the existing power connection.

Why the life cycle matters

The layered life cycle means data centers require continuous investment, not just a single construction bill. When companies discuss capital spending, part of it goes to new capacity and part to refreshing and upgrading what already exists. Keeping those two categories in mind helps make sense of why spending in this industry rarely stops, even when no new buildings are announced.

This article is general education about how the industry and financial reporting work. It is not investment advice and does not recommend buying or selling any security. For decisions about your own money, talk to a licensed financial advisor.

The number that gets misunderstood

Data-center companies often report capacity in megawatts. The number can describe a utility agreement, a customer contract, a building’s design capacity, a construction phase or a live load. These are not the same thing.

“Contracted MW” generally means capacity subject to an executed agreement. “Energized MW” means power has been delivered to the relevant electrical system and can serve a load, usually after interconnection, substation and commissioning steps. A project can have 400 MW contracted and 100 MW energized at the same time.

Applied Digital showed the distinction clearly in a 2026 investor presentation for its Polaris Forge 1 project: 400 MW contracted, 100 MW energized and 300 MW under construction. The numbers describe different stages of the same site, not a contradiction.

What contracted means

Contracted capacity can refer to a customer lease, a power-service agreement, a capacity reservation or another binding commitment. The exact definition depends on the company. A 2026 SEC filing described contracted power capacity as sellable power capacity subject to executed customer contracts, including capacity already generating revenue and capacity contracted but not yet in service.

A contract creates commercial visibility. It may support financing, construction planning and revenue forecasts. But it can still be conditional on permitting, customer deployment, construction, interconnection and acceptance.

The contract may also cover critical IT load rather than the total facility load. Critical IT load is the power available for servers, storage and network equipment. Cooling, lighting, pumps and building systems consume additional power. Two companies can report the same MW number while measuring different electrical boundaries.

What energized means

Energization is a physical milestone. It usually means a utility or on-site generator has energized the substation, switchgear or facility electrical system. It does not necessarily mean the data hall is fully commissioned or that customers are using the entire capacity.

A site can be energized in phases. The utility may energize one feeder or substation bay, allowing the first building to start commissioning while later phases remain under construction. The energized number can therefore be smaller than the contracted number and larger than the live IT load.

Core Scientific’s 2025 filing distinguishes gross utility power capacity, total leasable customer power capacity and other measures. It warns that these metrics are management estimates based on operational and engineering data and may not be comparable with similarly titled metrics from other operators.

The conversion chain

A typical project passes through a sequence:

1. Land and permits.
2. Utility study and capacity reservation.
3. Executed power or customer contract.
4. Equipment procurement and construction.
5. Interconnection approval.
6. Substation energization.
7. Data-hall commissioning.
8. Customer installation and load ramp.
9. Revenue-producing utilization.

At each stage, the risk changes. Contract risk concerns customer commitment and termination rights. Construction risk concerns labor, equipment and cost. Energization risk concerns utility work, protection testing and safety. Load-ramp risk concerns servers, networking, cooling and customer deployment.

How to read a project slide

Start with definitions and dates. Does the slide say “as of June 30” or “expected by 2028”? Is the metric gross capacity, critical IT load or customer demand? Does it include future phases, options or capacity that is not yet under construction?

Then reconcile the numbers. If contracted MW is higher than energized MW, ask what must happen to close the gap. If energized MW is higher than operating MW, ask whether the balance is commissioned, reserved for a customer or simply available.

The gap is not automatically a problem. A new campus may intentionally build electrical infrastructure ahead of tenant installation. But a large gap needs funding, equipment and a dated customer ramp.

The revenue connection

Contracted capacity is often used as a proxy for future revenue. That can be useful if the contract has a clear term, price, minimum commitment and commencement date. It is weaker if the customer can delay deployment, terminate, reduce load or avoid payments until acceptance.

Energized capacity is closer to an operational asset, but even it does not prove revenue. A facility can have power but no commissioned servers, no customer acceptance or no billable load.

BOSSTOX reporting discipline

BOSSTOX should report at least four numbers where available:

• Contracted customer IT load.
• Utility or site power capacity.
• Energized capacity.
• Live or revenue-producing IT load.

Each should include the date, definition and boundary. If the company gives only one number, quote the company’s wording and avoid translating it into a stronger claim.

BOSSTOX takeaway

Megawatts are not interchangeable. Contracted MW tells you what someone has committed to. Energized MW tells you what the electrical system has made physically available. Operating MW tells you what equipment is running. Revenue-producing MW tells you what is being billed.

The difference between those numbers is where a large part of data-center development risk lives.

Energization also has a technical meaning that should be stated. A utility may energize a substation, while the customer’s medium-voltage distribution remains locked out for testing. A site may have a live feeder but no commissioned UPS, cooling plant or generator synchronization. Conversely, a customer may run a temporary load before the permanent campus is complete. These cases make the phrase “energized” useful but incomplete.

The cleanest company disclosure presents a table with contracted, under construction, energized, commissioned and occupied MW. It should identify whether each figure is gross utility power, critical IT load or actual demand. Analysts can then calculate the conversion gap instead of relying on a single headline number.

A contract can also cover future customer capacity that has not yet been accepted. The company may describe the customer as committed while billing begins only after a hall is delivered. This is why contracted MW is useful for backlog and revenue visibility but should not be treated as current load. The gap can close quickly, or it can remain open through multiple quarters if construction, utility work or customer deployment slips.

The practical test is whether the next physical milestone has an owner and a date. If a project has a utility agreement but no transformer delivery date, the conversion risk is high. If the substation is energized but the customer has not installed equipment, the risk has moved from power to deployment.