Data Center Design Objective Inputs

The proposed design configuration—comprising 20% Air Cooled and 80% Liquid Cooled infrastructure, with allocations of 15% for AI Training off-takes, 50% for AI Inference off-takes, and 35% for Hyperscale Co-Location—creates a highly favorable revenue model. By strategically diversifying capacity across these key operational segments, the plan not only optimizes performance and energy efficiency but also positions the design for long-term growth and profitability.

Additionally, future data center capacity requirements have also been taken into consideration. During this process, when appointing the technical design engineering team—such as SUDLOWS—to carry out the work, the RFP should ensure that SUDLOWS leverages its network to bring tenants on board and provides an introduction email. This approach allows you, Dorian, Larry, and the rest of your team to focus on managing the technical qualification, while business development and scheduling of meetings are handled by other dedicated team members.

Additionally, you will need to collaborate with the Project Management Consultant (PMC), cost consultants, and vendors to obtain accurate cost structures, discounts, and guarantee bonds for their participation in the vendor selection process. These bonds can take the form of a Guarantee Bond or an Insurance Bond for the products—critical components of the project feasibility study and essential for applying any discounted rates. Given your overarching objective to optimize cost per kWh (especially in light of government incentives) and capitalize on efficiency-based discounts, securing these bonds and verifying vendor claims through case studies and recommendations will help ensure maximum profitability. Ultimately, the higher the overall efficiency of the data center, the greater the financial return.

Furthermore, the feasibility financial model should be aligned with regular operational updates, requiring weekly or biweekly reporting on the kWh cost reductions achieved by the appointed teams. This approach not only helps maintain alignment on efficiency and cost-savings objectives but also allows for ongoing adjustments to strengthen the data center’s overall viability. Additionally, integrating decarbonization strategies—such as renewable power sourcing and on-site power generation—and exploring additional power generation options will further optimize operational costs, reduce carbon emissions, and enhance the long-term sustainability of the data center.

Below is a high-level overview of the major data center systems that significantly impact energy usage, along with guidance on the efficiencies vendors typically need to offer. Keep in mind that reducing your actual cost from $0.17 per kWh to $0.01 per kWh is typically not feasible by equipment efficiency alone, because the utility’s per-kWh rate is usually fixed and outside your control. Instead, you focus on reducing total consumption so that your data center’s effective spending per unit of compute approaches that target.

1. Major Equipment Categories for Power Efficiency

1. IT / Server Hardware

   • Key Efficiency Factors: CPU/GPU power efficiency, power supply efficiency (e.g., 80 PLUS Platinum or Titanium), dense virtualization, low-power idle modes.

   • Efficiency Targets:

     • Power supply efficiency over 96% for most load ranges.

     • Advanced power management features to reduce consumption at low loads.

   • Potential Reduction in IT Load: 15–30% improvement over standard hardware.
     

2. Cooling & HVAC Systems

   • Key Efficiency Factors:

     • Cooling approach (free cooling, liquid cooling, hot/cold aisle containment),

     • Variable-speed fans and chillers,

     • Adiabatic/evaporative cooling where climate-appropriate.

   • Efficiency Targets:

     • Drop mechanical cooling energy usage by 30–50% through containment and free cooling.

     • Achieve a lower Power Usage Effectiveness (PUE)—for example, aiming for <1.2 in many modern data centers.

   • Potential Reduction in Overall Data Center Power: 20–40% of overhead power, depending on baseline.
     

3. Uninterruptible Power Supply (UPS) & Power Distribution

   • Key Efficiency Factors:

     • High-efficiency UPS topology (e.g., transformerless UPS, modular UPS systems with >95% efficiency),

     • Lower-loss power distribution units (PDUs).

   • Efficiency Targets:

     • 97–99% UPS efficiency during normal operation.

     • Avoid double-conversion losses when not required.

   • Potential Reduction in Power Losses: 5–10% of total facility power, depending on baseline.
     

4. Lighting and Auxiliary Systems

   • Key Efficiency Factors: LED lighting, motion sensors, intelligent controls for lighting in seldom-accessed areas.

   • Efficiency Targets: 70–80% reduction in lighting energy use relative to legacy fluorescent solutions.

   • Potential Reduction in Overall Data Center Power: Usually small (1–2%), but still worth including.
     

5. Building Management System (BMS) / Data Center Infrastructure Management (DCIM)

   • Key Efficiency Factors:

     • Monitoring and controlling temperature, humidity, airflow,

     • Automated scheduling of cooling equipment based on real-time load.

   • Efficiency Targets: Overall 5–15% energy savings through optimized operation and identifying hotspots/waste.
     

2. Approaching the “$0.01/kWh” Effective Cost Goal

Again, the utility rate itself (17¢/kWh) is largely fixed; however, if you can reduce total consumption enough, you bring down your cost of operation per unit of compute. In other words:

Effective cost per compute unit=(kWh used) × (Cost per kWh)Compute OutputEffective cost per compute unit=Compute Output(kWh used) × (Cost per kWh)​ If you reduce your total energy use by ~94% (while delivering the same compute services), you effectively reduce your cost per unit of compute to near $0.01/kWh in practice, even if the nominal utility rate remains $0.17/kWh.

Illustrative Breakdown of Efficiency Improvements

Below is an example breakdown showing how multiple improvements could compound toward an ~80–90% overall usage drop from a typical, unoptimized baseline data center. (Exact results vary widely by site, climate, baseline technology, and operational practices.)

1. Advanced Server Hardware & Virtualization:

   • 25–30% reduction in total power (due to lower CPU/GPU wattage, better PSUs, and higher compute density).

2. State-of-the-Art Cooling Design (containment, free cooling, liquid cooling):

   • 30–50% reduction in cooling power (relative to legacy designs).

3. High-Efficiency UPS & Power Distribution:

   • 5–10% reduction in facility-side losses.

4. Optimized BMS/DCIM:

   • 5–15% overall energy savings from real-time controls, reduced waste, and automated scheduling.

5. Lighting & Other Optimizations:

   • 1–2% overall, but still contributes incrementally.
     

When combined, these strategies can approach a total overhead reduction of 60–70% (or higher) from a typical older data center environment. Achieving a 90%+ reduction requires extreme optimization—plus potentially on-site generation (e.g., solar, fuel cells) or negotiated utility rates—so treat that as a best-case scenario rather than a baseline expectation.

3. Setting Vendor Efficiency Criteria

When creating your vendor list or awarding a contract, you might:

1. Specify Minimum Efficiency Levels

   • Example: Servers must use 80 PLUS Platinum/Titanium-certified PSUs.

   • UPS systems must maintain at least 95–97% efficiency at 50–100% load.

   • CRAC (computer room air conditioning) units must have variable-speed drives, meet or exceed certain ASHRAE recommendations, etc.

2. Require PUE and Cooling Targets

   • For instance, vendors must demonstrate that their recommended cooling system design can achieve a PUE of <1.3 or even <1.2 in your region’s climate.

3. Demand Detailed Power Modeling

   • Insist on measurement-based data or computational fluid dynamics (CFD) modeling to show exactly how the vendor’s solution lowers consumption—rather than just a marketing claim.

4. Incorporate Penalties/Rewards in Contracts

   • Align incentives: if the design exceeds the targeted PUE or consumes more kWh than stated, the vendor faces penalties; if it meets or beats targets, they earn bonuses.
     

4. Putting It All Together

• Goal: Dramatically reduce total energy consumption so the operational cost per unit of compute approaches the equivalent of $0.01/kWh.
  

• Reality: True cost per kWh from the utility is rarely negotiable down from $0.17 to $0.01—so the data center must focus on:

  • Best-in-class IT equipment efficiency,

  • Highly optimized cooling,

  • Low-loss power infrastructure,

  • Rigorous monitoring and automation,

  • (Optionally) on-site generation to offset or partially replace expensive grid power.
    

This holistic approach—plus the right contractual terms with vendors—creates the path toward the effective cost reduction you want, even though the utility rate may remain at 17¢/kWh on paper.

Notes

1. Combining Optimizations

   • The cumulative effect of improvements in server efficiency, cooling, UPS, and infrastructure design can reduce total facility power usage by 50–70% (or more) compared to a legacy data center.

   • Extreme optimization (plus on-site generation or highly discounted utility rates) is often required to achieve an effective $0.01/kWh when the nominal cost is $0.17/kWh.

2. Contract Criteria

   • When creating an RFP (Request for Proposal) or awarding contracts, consider specifying minimum efficiency standards (e.g., 80 PLUS Titanium for power supplies, UPS >95% efficiency, PUE <1.3, etc.) and requiring proof of performance via modeling, testing, or real-world metrics.

   • Tie payment incentives or penalties to actual performance (e.g., measured PUE, UPS loss, cooling load) to ensure vendors deliver on efficiency promises.

3. Effective vs. Actual kWh Cost

   • Because the utility rate (17¢/kWh) is typically fixed, the effective cost per kWh of delivered compute or service declines when total consumption is dramatically reduced.