Keentel Engineering Data Center Design Services: Designing for Uptime, Grid Reliability, and Scalable Growth

The data center industry has entered a different era. What used to be treated as a high-reliability commercial facility is now increasingly viewed as an emerging large load with direct implications for transmission planning, grid operations, system stability, power quality, and long-term resource adequacy. NERC’s July 2025 white paper makes that shift unmistakable. It explains that data centers are among the fastest-growing energy consumers in North America, and it notes forecasts that data centers alone may account for as much as 12% of all U.S. electricity consumption by 2028, up from 4.4% in 2023. 

For owners, developers, hyperscalers, utilities, and EPC teams, this means one thing: data center electrical design can no longer stop at the building fence. The facility must be engineered as part of a larger electrical ecosystem. That is where Keentel Engineering’s data center design services can create real value by helping clients design facilities that do not just achieve uptime targets, but also satisfy interconnection expectations, reduce operational surprises, and support reliable long-term expansion. This service positioning is a practical engineering response to the risks and design implications described throughout the NERC paper.

Why modern data center design is different

NERC defines large loads as commercial or industrial facilities, or aggregations of load at a single site, that can create reliability risks due to demand, operational characteristics, or other factors. Importantly, the paper does not reduce the issue to a single MW threshold. It emphasizes that while many industry participants mentioned 50 MW or 75 MW as useful reference points, a meaningful large-load definition must also account for factors such as ramp rate, real-time behavior, flexibility, protection systems, backup power schemes, voltage sensitivity, and interconnection context. That is a major insight for data center developers: the engineering challenge is not just how big the facility is, but how it behaves electrically.

That behavior matters because modern data centers are not simple static loads. NERC describes data centers and other computational loads as power-electronic-load-heavy facilities with high energy consumption, variable operational demand, significant cooling requirements, internal protection logic, and backup power systems. Some data center loads are pulsed and non-linear, with extremely fast ramping characteristics that can introduce both stability and power quality concerns. In other words, the electrical design problem is no longer just service entrance sizing and generator redundancy. It is now a matter of dynamic system integration. 

The four building blocks that shape data center electrical design

The white paper identifies four main data center components:
IT-related equipment, power delivery systems, cooling systems, and miscellaneous lighting/security loads. It further notes that IT-related equipment may represent 60% to 95% of total facility demand, while cooling systems commonly rely on power electronics such as variable-speed drives and inverters that can create harmonics and increase reactive power demand. It also highlights that AI-oriented HPC facilities may use different power continuity strategies than traditional data centers, sometimes favoring checkpoint recovery rather than full UPS coverage for all IT equipment. 

For Keentel Engineering, this points to a strong service message: successful data center design requires coordinated engineering across medium-voltage distribution, backup generation, UPS strategy, low-voltage distribution, thermal plant integration, harmonic mitigation, grounding, protection coordination, and utility-facing interconnection design. If those pieces are engineered separately, the owner may end up with a facility that functions internally but creates unnecessary risk at the point of interconnection. The better approach is an integrated design philosophy where building power architecture and grid-facing behavior are developed together. That is the practical takeaway from the paper’s description of modern data center configuration and risk.

Traditional data centers are not the same as AI training facilities

One of the most important insights in the NERC paper is that “data center” is too broad a term to support one-size-fits-all engineering. Traditional data centers have historically been smaller, often under 30 MW, with high redundancy and relatively limited variability. By contrast, AI training data centers may exhibit sharp load changes tied to training runs and checkpoint events. NERC cites an example in which a 50 MW block of a larger 200 MW AI training data center changed demand at a rate of 1.9 p.u. per second for about 250 milliseconds, and it notes that transitions between training and checkpoint saving can occur in under one second. AI inference facilities may still be high-power facilities, but current inference methods do not necessarily show the same rapid ramping pattern observed in training loads. 

This distinction matters because owners often ask for “data center electrical design” as if every facility has the same load signature. It does not. Keentel Engineering can differentiate itself by framing its design process around workload-informed electrical engineering. A cloud facility with steady utilization should not be studied exactly the same way as an AI training campus with rapid transitions, nor should either be treated like a behind-the-meter hybrid load with co-located generation. The NERC paper strongly supports a design philosophy based on facility-specific electrical characterization, not generic templates.

Grid interconnection is now a design issue, not a late-stage paperwork issue

Another major lesson from the white paper is that utilities, ISOs, and transmission planners are being pushed to revisit how they handle large load interconnections. NERC documents existing constructs such as ERCOT’s 75 MW threshold for large loads and Dominion Energy’s 100 MW transmission-tap threshold, along with Dominion’s planning limits such as ring-bus expectations above 100 MW, substation loading constraints, and information requirements tied to ride-through capabilities. These examples show that the utility side of the project is becoming more structured, more technical, and more dependent on customer-provided data. 

For Keentel Engineering, this creates a clear market position: help clients arrive at interconnection discussions with the right technical package the first time. That means developing conceptual one-lines, demand block definitions, staged energization strategy, ride-through expectations, backup system operating philosophy, load segmentation logic, and the necessary studies that demonstrate the project can be connected reliably. The paper repeatedly shows that missing models, unclear protection behavior, and insufficient visibility into load characteristics can create planning and operating risks. In that environment, engineering firms that can translate owner intent into a utility-ready technical submittal will be increasingly valuable.

Observability and modeling are now part of good design

The paper states that operators and planners need accurate information on interconnection timelines, peak demand, load behaviors, protection and control settings, and dynamic models in order to study large loads properly. It also warns that lack of high-speed recording data, PMUs, and other monitoring can make root-cause analysis difficult or impossible after an event. NERC further notes that large loads are not required to register as NERC entities under current criteria, which can make it harder for system operators to obtain timely information after disturbances.

That means a modern data center design scope should go beyond wires and breakers. Keentel Engineering can position observability as a core service deliverable: telemetry architecture, metering philosophy, high-resolution disturbance recording, event visibility, and model-ready data collection. Owners often think of these items as operational extras. The NERC paper suggests they are increasingly part of the reliability foundation. A well-designed facility should not only consume power safely; it should also make its behavior understandable to the owner, operator, utility, and planning teams.

Planning for ramping, reserves, and forecasting

One of the strongest sections in the white paper concerns operations and balancing. NERC explains that large loads, especially power-electronic-rich loads, can change consumption in seconds, much faster than conventional generators can ramp. It gives a data center example in which load dropped from roughly 450 MW to 40 MW in 36 seconds, remained near 7 MW for about four hours, and later returned to 450 MW within minutes. NERC warns that such rapid shifts can challenge reserve procurement, frequency regulation, and voltage control. It also notes that many large loads do not submit real-time or day-ahead operating profiles, which degrades forecasting accuracy. 

For data center developers, this means the load profile itself is now an engineering deliverable. Keentel Engineering can create value by helping owners characterize expected ramp behavior, identify load blocks, define operating modes, and coordinate expected transitions with utilities and balancing authorities where relevant. This is especially important for AI facilities, phased campuses, and sites with flexible computational strategies. A project that cannot explain how it ramps, recovers, sheds, or restarts is a harder project to interconnect and a riskier one to operate. The paper makes clear that forecasting and reserve challenges are not abstract concerns; they are direct consequences of poor visibility into load behavior.

Stability and ride-through cannot be left to chance

NERC identifies several high-priority risks associated with emerging large loads, including resource adequacy, balancing and reserves, ride-through, voltage stability, angular stability, and oscillations. The ride-through discussion is especially important for data centers. The paper notes that some facilities may switch to backup systems after multiple transient voltage disturbances in a short period and cites real-world events in which roughly 1,500 MW of voltage-sensitive load, primarily from data centers, was lost following transmission faults. NERC also stresses that many stability concerns arise because loads have not been modeled accurately enough in planning studies.

This is a major opportunity for Keentel Engineering’s service messaging. Data center electrical design should include explicit attention to ride-through philosophy, transfer logic, protection coordination, backup generation sequencing, voltage sensitivity, reconnection behavior, and dynamic study support. In the old model, the goal was often to keep the data center itself alive. In the new model, the facility must also avoid becoming a source of sudden, system-wide disturbance. The strongest engineering service providers will be the ones who help owners satisfy both objectives at once.

Harmonics, flicker, and reactive behavior are not secondary issues

NERC’s power quality discussion confirms what many experienced engineers already suspect: modern data centers are major users of power electronics on both the IT side and the cooling side. The paper explains that these devices can make data centers significant sources of harmonics, especially if filtering is not designed deliberately. It also warns that transitions to higher-power pulses can create voltage fluctuations, flicker, unbalance, and broader power quality concerns. In one cited case, voltage distortion was significantly reduced after harmonic mitigation measures were implemented. The paper also notes reported oscillation behavior associated with data center UPS input units in at least one system.

For Keentel Engineering this supports a data center design offering that includes harmonic studies, reactive power assessment, filter evaluation, grounding and bonding review, VFD/UPS interaction analysis, and power quality mitigation strategy. Owners often focus on availability, but a facility that produces power quality problems may face utility pushback, nuisance trips, thermal stress, and long-term operational friction. Harmonic and power quality engineering should be treated as part of the core design package, not as a corrective measure after commissioning problems appear.

Behind-the-meter and co-located generation add another layer of complexity

The white paper devotes attention to the difference between front-of-meter and behind-the-meter large load configurations. It explains that behind-the-meter large loads may have less operational visibility and that unexpected transitions from co-located generation to grid supply can create sudden demand spikes and stability concerns. It emphasizes the need for clear standards, monitoring, and coordination to prevent unplanned swings in demand. 



That is especially relevant as more data center projects explore gas generation, renewable integration, hybrid architectures, and island-capable strategies. Keentel Engineering can strengthen its value proposition by offering design support for front-of-meter/behind-the-meter architecture evaluation, transition studies, protection schemes, synchronization logic, backup and primary-source operating modes, and utility coordination for hybrid sites. The NERC paper suggests that these configurations must be engineered with disciplined attention to failure modes, because the grid impact of a transition event may be much larger than the owner first assumes.

Restoration and segmentation should be designed before the emergency happens

A particularly underappreciated insight in the white paper is the importance of load segmentation during system restoration. NERC explains that large loads may need to be restored in smaller, manageable, predictable blocks and warns that unclear segmentation can lead to frequency decline, voltage collapse, unintended UFLS operation, or the need for additional load shedding. It also notes that as large loads grow, UFLS obligations and manual load-shed planning may need to be revisited more often than traditional review cycles assumed.

This means staged energization and restoration logic should be built into the design, not left to operations manuals after construction. Keentel Engineering can translate this into tangible deliverables: segmented load block planning, restoration sequences, blackstart-aware re-energization philosophy, generator pickup logic, and load-shed coordination concepts. Owners care deeply about how fast they can return to service after a disturbance. Utilities care deeply about whether that return happens in a stable and predictable way. Good engineering has to satisfy both.

What Keentel Engineering’s data center design services should emphasize

Based on the NERC paper, the strongest technical message for Keentel Engineering is that data center design must be approached as a whole-system electrical engineering problem. That means combining:

• utility and interconnection readiness, 
• medium- and low-voltage electrical system design, 
• backup generation and UPS architecture, 
• ride-through and protection coordination, 
• load characterization and dynamic behavior review, 
• harmonics and reactive power mitigation, 
• telemetry and observability design, 
• phased energization and restoration planning, 
• and study support for reliability-driven stakeholder review. 

This service framing is an engineering inference drawn from the risks, characteristics, and recommendations presented in the NERC white paper. The paper explicitly calls for better classification of large loads, improved mitigation approaches, better load models, stronger understanding of protection impacts, and improved methods for assessing resource adequacy risks. A data center engineering consultant that helps clients address those needs early is well aligned with where the industry is heading.

Final thought

The future of data center design belongs to engineering teams that understand both facility resilience and grid consequences. NERC’s July 2025 white paper makes clear that modern data centers—especially AI-oriented facilities—are no longer passive loads. They are electrically dynamic, operationally consequential, and increasingly central to transmission and reliability planning. For Keentel Engineering, that creates a powerful market position: help clients build data centers that are not only scalable and reliable inside the fence, but also credible, stable, and utility-ready at the point of interconnection.

Detailed FAQ Draft

Frequently Asked Questions About Keentel Engineering’s Data Center Design Services