BLOG

My perception of the world has always been through the lens of technical figures and financial implications. This analysis was directly inspired by a compelling article from Stanisław Kowalik, which truly resonated with my professional interest.

I wholeheartedly recommend its thorough reading: https://www.linkedin.com/feed/update/urn:li:activity:7346064934738223104/.

The developments with these turbines, as described by Stanisław, represent the next stage in this evolution, and are indeed a crucial advancement for semi-autonomous, local energy systems. His insights prompt reflection on the evolving landscape of energy.

My conceptual work with Stanisław Kowalik on semi-autonomous energy systems has long emphasized a paradigm where the traditional distribution grid serves as a crucial, yet supplementary, backup rather than the primary source. This conceptual framework underpins our approach to ensuring high reliability and economic efficiency. Furthermore, practical experience reinforces these principles. For example, my team have achieved significant successes in implementing generation systems based on a patented dual-fuel installation. This technology allows operation on diesel fuel combined with any gas, creating a unique blend of reliability and economy.

It is this blend of theoretical insight and practical success that frames the current TCO analysis.

For this TCO analysis, I selected the core of each power system:

• Gas Microturbine. FlexEnergy 250 kW, LNG-fueled, with integrated heat and cold recovery mentioned by Stanislaw.

• Diesel Generator. 250 kW industrial unit, with heat recovery.

• Grid Power. Energy purchased directly from the grid.

I used averaged market data for Poland. Fuel prices are based on recent data from July 2025.

1. Key Shared Assumptions

• Location. Poland.

• Target Facility. An AI-specialized server farm. It demands stable power, significant cooling, and potential heating (e.g., for offices or auxiliary facilities).

• Cost of electricity from the grid. 0.80 PLN/kWh (averaged for large consumers).

• Cost of LNG. 0.30 PLN/kWh (This specific value for LNG is adopted for this TCO analysis, considering regasification and transport costs, and is distinct from general natural gas prices in the provided source).

• Cost of diesel fuel (ON). 4.15 PLN/liter (average retail price with delivery, based on 4.05–4.26 PLN/l range)1.

• Cost of purchased heat (from grid/traditional). 0.15 PLN/kWh.

• Cost of purchased cooling (from grid/traditional). 0.20 PLN/kWh.

• Installation costs. 15% of equipment purchase cost.

• Disposal costs. 1% of purchase price (microturbine), 3% of purchase price (Diesel).

• Annual operating hours. 8000 hours/year (continuous operation).

• Server farm cooling demand. Approximately 100-150 kW of cooling. I have adopted 120 kW of continuous demand.

• Heating demand. I have adopted 80 kW of continuous demand (e.g., for offices or auxiliary facility heating).

2. Detailed Assumptions for Each System

2.1. FlexEnergy 250 kW Microturbine (LNG with heat and cold recovery - trigeneration)

• Electrical power. 250 kW.

• Electrical efficiency. 33%.

• Heat recovery. Typically 50-60% of fuel power. I adopted 55% heat recovery.

• Annual fuel consumption (kWh). (250 kW / 0.33) * 8000 h = 6,060,606 kWh/year.

• Annual heat production. 6,060,606 kWh * 0.55 = 3,333,333 kWh heat/year.

• Value of recovered heat (savings). 3,333,333 kWh * 0.15 PLN/kWh = 500,000 PLN/year.

• Cold recovery from LNG regasification. As previously noted, approx. 0.15 kWh of cold per 1 kWh of fuel.

• Annual cold production. 6,060,606 kWh * 0.15 = 909,091 kWh cold/year.

• Value of recovered cold (savings). 909,091 kWh * 0.20 PLN/kWh = 181,818 PLN/year.

• Purchase cost (estimated, with regasification, heat, and cold recovery systems). 1,800,000 PLN.

• Operational lifespan. 18 years.

• Annual O&M costs. 2.5% of purchase cost = 45,000 PLN/year.

• Annual fuel costs (LNG). 6,060,606 kWh * 0.30 PLN/kWh = 1,818,182 PLN/year.

• Emissions. Low.

2.2. Diesel 250 kW Generator (with Heat Recovery - Cogeneration)

• Electrical power. This Diesel generator is assumed to cover the total electrical demand of the facility, which includes 250 kW for the server farm IT load and an additional 40 kW for cooling, totaling 290 kW of electrical output.

• Electrical efficiency. 38%.

• Heat recovery. Typically 40-50% of fuel power. I adopted 45% heat recovery.

• Annual fuel consumption (kWh). (290 kW / 0.38) * 8000 h = 6,105,263 kWh/year.

• Annual heat production. 6,105,263 kWh * 0.45 = 2,747,368 kWh heat/year.

• Value of recovered heat (savings). 2,747,368 kWh * 0.15 PLN/kWh = 412,105 PLN/year.

• Purchase cost (estimated for a ~290-300 kW continuous duty generator with heat recovery). 850,000 PLN (previously 700,000 PLN).

• Operational lifespan. 12 years.

• Annual O&M costs. 5% of purchase cost = 42,500 PLN/year (previously 35,000 PLN).

• Annual fuel costs (ON). 6,105,263 kWh / 10 kWh/liter * 4.15 PLN/liter = 2,533,639 PLN/year.

• Fuel storage costs. 10,000 PLN/year.

• Oil/filter disposal costs. 4,000 PLN/year.

• Emissions. Higher. Potential emission fees. 5,000 PLN/year.

• Annual Cooling Costs (purchased from grid). 0 PLN (as the Diesel generator produces its own power for cooling).

• Annual Grid Electricity Costs. 0 PLN (as the Diesel generator covers the primary server farm load).

2.3. Grid Power (Main Source)

• Electrical power drawn. 250 kW.

• Equipment purchase cost. 0 PLN.

• Operational lifespan. 18 years (for lifecycle comparison).

• Annual O&M costs. 0 PLN.

• Fuel costs. N/A.

• Annual electricity costs. 2,000,000 kWh * 1.15 PLN/kWh = 2,300,000 PLN/year.

• Heating costs. Purchased from the grid. I adopted 80 kW of heating demand.

• Annual heat consumption. 80 kW * 8000 h = 640,000 kWh/year.

• Cost. 640,000 kWh * 0.15 PLN/kWh = 96,000 PLN/year.

• Cooling costs for server farm. As for Diesel. Annual. 320,000 kWh * 1.15 PLN/kWh = 368,000 PLN/year.

• Outage/interruption costs. High for AI server farms. UPS or backup generator required (not included in this TCO comparison).

• Emissions. Indirect.

3. Total Cost of Ownership (TCO) analysis - annual and total lifecycle comparison

TCO = Purchase Cost + Installation Costs + Sum of Annual Operating Costs * Operational Lifespan + End-of-Life/Disposal Costs

3.1. System Comparison for AI Server Farms (250 kW Power)

3.2. Assessment of TCO for AI Server Farms (Trigeneration vs. Cogeneration vs. Grid):

• LNG Microturbine with heat and cold recovery (FlexEnergy 250 kW). This system yields the lowest TCO. Its higher initial purchase cost is offset by significantly lower fuel costs and substantial savings from full energy utilization. This includes electricity, heat, and notably, free cold from LNG regasification, which is critical for server farms. This represents a comprehensive systemic solution.

• Diesel 250 kW (Generator with heat recovery). It presents the lowest purchase cost but incurs a higher TCO than the LNG microturbine. Fuel costs, a shorter lifecycle, and the necessity of purchasing cooling (due to lack of cold recovery) make this a less economical primary power source for an AI server farm. Its function remains largely confined to backup operations.

• Grid Power. Despite zero initial investment costs, it is significantly more expensive in the long term than both the LNG microturbine and the Diesel generator. The high costs of purchased electricity, heat, and cooling accumulate, rendering this solution financially inefficient.

4. Key Decision Factors (Beyond TCO)

My analysis of energy solutions extends beyond mere TCO figures. Here are additional, pivotal aspects for AI server farms.

Simplified cost-effectiveness comparison of three power supply systems for an ai data center

• Energy Synergy (Trigeneration). LNG microturbines achieve the highest overall efficiency. This is due to simultaneous electricity, heat, and cold production. AI server farms represent ideal environments for trigeneration. Cold recovery from LNG is a transformative element. It obviates the need for external chillers. This optimizes the server farm's Power Usage Effectiveness (PUE).

• Reliability and Power Continuity (Uptime). AI server farms demand near 100% uptime. Local generation from a microturbine ensures stable power delivery. It reduces reliance on potentially unstable grid supply. Diesel remains an excellent, rapid backup. It is not a primary operational solution.

• Environmental Impact and Regulations. LNG microturbines are the cleanest among combustion-based solutions. They exhibit low emissions. This is crucial for sustainable development goals and adherence to future, stricter environmental regulations. Diesel exhibits high emissions. Escalating emission fees and public expectations will substantially increase its hidden costs.

• Noise and Vibrations. Microturbines operate quietly. This minimizes environmental disturbance and reduces costs associated with acoustic insulation. Diesel generators are loud and vibrate. They necessitate significant investment in soundproofing and structural reinforcement.

• Fuel Safety. Natural gas/LNG offers safer storage compared to diesel fuel. It eliminates risks associated with leaks and ground contamination.

5. Instead of the summary - concluding remarks

This cautious TCO analysis appears to offer a definitive conclusion – but assumptions are always open for discussion. For AI-specialized server farms, an LNG-fueled microturbine, incorporating heat and cold recovery (trigeneration), presents itself as the most probable optimal choice. It is poised to deliver the lowest TCO, the highest energy efficiency, reliable power supply, and full alignment with sustainable development objectives.

While the initial investment is higher, the long-term savings, synergy with server farm cooling, and environmental benefits establish it as a rational and forward-looking decision.

Hence, my comment on Stanisław Kowalik's post suggesting that we are likely witnessing a subtle, yet significant, breakthrough.