Data Center Fundamentals·Cooling & HVAC Systems

Free Cooling & Economization

Learn how data centers use outside air to reduce cooling costs and improve PUE.

Intermediate12 min readLesson 15 of 31

Introduction Data centers consumed 460 TWh of electricity globally in 2022, with cooling systems accounting for 30-40% of that total energy use.

That translates to roughly $40 billion in annual cooling costs across the industry.

Free cooling and economization represent the most significant opportunity to reduce this burden-facilities implementing these technologies report PUE improvements from 1.6-1.8 down to 1.2-1.3, cutting cooling energy consumption by 50-75% during favorable conditions.

Free cooling leverages naturally cold outdoor air or water to dissipate heat, eliminating or reducing mechanical refrigeration loads.

Economization is the control strategy that enables facilities to transition between mechanical cooling and free cooling modes based on outdoor conditions.

When Facebook (now Meta) designed its Prineville, Oregon facility in 2011, free cooling was still considered experimental at scale.

Today, every major hyperscaler mandates economization in new builds, and retrofit projects dominate colocation providers' capital expenditure priorities.

This lesson breaks down the two primary economization approaches-airside and waterside-examines the climate factors that determine economic viability, and provides frameworks for calculating potential savings at your facility.

You'll understand why Google targets locations with 350+ annual hours below 50°F for new data centers, and why Digital Realty invested $180 million upgrading existing facilities with economizer systems between 2019-2022.

Airside Economization: Direct and Indirect Approaches Airside economization uses outdoor air to cool data center spaces, either by bringing it directly into the facility or using it to cool air in a heat exchanger.

The fundamental physics are simple: when outdoor air temperature drops below the return air temperature from the data hall (typically 75-85°F), mechanical cooling becomes partially or completely unnecessary. Direct airside economization introduces filtered outdoor air straight into the data hall.

Microsoft's Quincy, Washington facilities pioneered this approach at hyperscale, operating without chillers for approximately 4,500 hours annually (51% of the year).

The facility uses 40-foot-tall air handling units that can process 2 million cubic feet per minute.

When outdoor temperatures fall between 35-50°F, operators achieve PUE values as low as 1.06.

Direct economization requires minimal equipment-sophisticated filtration systems, dampers, and controls-making it the lowest capital cost option at approximately $150-200 per kW of IT capacity.

The tradeoff comes in contamination risk.

Bringing unprocessed air into a facility introduces particulate matter, humidity variation, and potentially corrosive gases.

ASHRAE's 2021 Particulate and Gaseous Contamination Guidelines establish acceptable thresholds, but real-world conditions vary dramatically.

Switch's Tahoe Reno facility addresses this with MERV-14 filtration and continuous air quality monitoring, replacing filters every 30-45 days instead of the 90-day cycles typical for recirculation systems. Indirect airside economization uses a heat exchanger (air-to-air heat wheel or plate-and-frame unit) to transfer cooling from outdoor air to facility air without direct mixing.

Google Cloud's Belgian facilities employ this approach, achieving 95% free cooling hours annually while maintaining strict contamination controls.

The capital cost runs higher-$300-400 per kW-but operational risk drops substantially.

Airflow rates on the outdoor side can exceed indoor rates by 20-30%, maximizing heat transfer while keeping contaminated air isolated from IT equipment.

Here's how the two approaches compare across key metrics:

Metric Direct Airside Indirect Airside
Capital Cost (per kW) $150-200 $300-400
Free Cooling Hours (Chicago climate) 3,200-3,800 2,800-3,400
Contamination Risk High Low
Humidity Control Complexity High Medium
Maintenance Hours (annual) 400-600 200-300
PUE at Full Economization 1.05-1.10 1.08-1.12

Waterside Economization: Closed Circuit Efficiency Waterside economization cools facility water (used in CRAC/CRAH units or rear-door heat exchangers) through evaporative cooling towers or dry coolers, bypassing mechanical chillers.

Rather than cooling air directly, this approach leverages water's superior thermal capacity-approximately 3,500 times that of air-to transport heat from IT equipment to outdoor rejection points.

Traditional chilled water systems maintain supply temperatures of 42-48°F year-round, requiring constant compressor operation regardless of outdoor conditions.

Waterside economizers allow supply temperatures to float upward as outdoor wet-bulb temperatures decrease.

When wet-bulb temperatures drop below 45-50°F, many facilities achieve 100% free cooling.

Digital Realty's Chicago campus operates chillers for fewer than 400 hours annually using this approach, down from 6,500 hours pre-retrofit.

The key performance factor is wet-bulb temperature, not dry-bulb.

Wet-bulb represents the lowest temperature achievable through evaporative cooling-essentially the combination of temperature and humidity.

A 70°F day with 30% relative humidity delivers a 50°F wet-bulb temperature, enabling full free cooling.

The same 70°F day at 80% humidity produces a 66°F wet-bulb, requiring partial mechanical assistance.

Equinix's Dallas data centers demonstrate waterside economization in challenging climates.

Dallas averages only 850 hours annually below 45°F wet-bulb, yet the facility still achieves 35% free cooling hours through aggressive temperature setpoint management.

Return water temperatures float from 60-72°F (instead of the traditional fixed 58°F), allowing economizer operation at higher outdoor temperatures.

This strategy requires IT equipment rated for higher inlet temperatures-typically 80-85°F versus the legacy 68-72°F standard-but ASHRAE's expanded thermal guidelines (A1-A4 classifications) now accommodate these conditions.

CyrusOne's Phoenix facilities take waterside economization a step further with hybrid dry cooler systems.

Traditional cooling towers require continuous water consumption-approximately 1.8 liters per kWh of heat rejection.

Dry coolers use only ambient air, eliminating water usage but requiring larger surface areas and fan power.

The hybrid approach switches between modes based on outdoor conditions: dry cooling during Phoenix's 4,000+ hours below 60°F, evaporative assist during extreme heat.

Annual water usage drops from 120 million gallons (traditional cooling towers) to 22 million gallons (hybrid), while maintaining PUE below 1.25.

Climate Zone Analysis and Economic Viability Free cooling effectiveness correlates directly with climate, but the relationship isn't linear.

A facility in Minneapolis (6,200 heating degree days, cold continental climate) doesn't automatically outperform one in San Francisco (2,800 heating degree days, Mediterranean climate).

The distribution of temperatures throughout the year matters more than simple averages.

ASHRAE climate zones provide the foundational framework.

Zones 1-3 represent hot climates with limited economization potential, zones 4-5 offer moderate opportunities, and zones 6-8 deliver maximum benefit.

Here's what the data shows for annual free cooling hours using indirect airside economization with a 55°F changeover setpoint:

Climate Zone Representative City Annual Free Cooling Hours Percentage of Year Estimated PUE Improvement
2A Houston, TX 850 10% 0.02-0.04
3A Atlanta, GA 2,100 24% 0.08-0.12
4A Chicago, IL 3,400 39% 0.15-0.22
5A Boston, MA 4,200 48% 0.20-0.28
5B Denver, CO 5,100 58% 0.25-0.35
6B Helena, MT 5,800 66% 0.30-0.40
8 Fairbanks, AK 7,400 84% 0.40-0.55

Of the company's 30 availability zones announced between 2018-2023, 73% are located in climate zones 4-6, despite these regions representing only 35% of US population centers.

The Oregon, Ohio, and Virginia regions leverage 4,500-5,000 annual free cooling hours, reducing operational expenses by $12-15 million annually per 50 MW facility compared to equivalent capacity in hot climates.

But climate zone alone doesn't tell the complete story.

Humidity, altitude, and diurnal temperature swings all impact performance.

QTS's Irving, Texas facility (climate zone 3A) achieves 3,200 free cooling hours despite being in a challenging climate.

The key? Dallas-Fort Worth's 18-22°F average diurnal temperature swing means even 95°F summer days drop to 73-75°F overnight.

Running compute workloads during these overnight windows reduces cooling loads during peak temperature hours, effectively increasing economization time.

Altitude provides another advantage through reduced air density and improved evaporative cooling performance.

CoreSite's Denver facilities operate at 5,280 feet elevation, where boiling point drops to 203°F and evaporative cooling efficiency increases by approximately 15%.

The facility achieves waterside economization for 5,600 hours annually with wet-bulb setpoints 3-4°F higher than sea-level equivalents.

Control Strategies and Integration Complexity Economization sounds straightforward in theory-when it's cold outside, use outdoor conditions for cooling.

Implementation requires sophisticated control systems that balance multiple competing objectives: maintaining tight temperature and humidity tolerances, preventing rapid cycling between cooling modes, managing contamination risk, and optimizing energy consumption across interdependent systems.

The changeover strategy determines when facilities transition between economization and mechanical cooling.

Fixed setpoint approaches switch at predetermined outdoor temperatures (typically 50-55°F dry-bulb for airside, 45-50°F wet-bulb for waterside).

This simplicity aids operational predictability but leaves efficiency on the table.

Google's machine learning-optimized systems use dynamic setpoints that consider 27 variables including forecast data, compute load predictions, equipment efficiency curves at partial loads, and electricity pricing.

Meta's Prineville facility runs integrated economizer control that adjusts every 60 seconds.

Outdoor air dampers modulate continuously rather than operating binary open/closed, creating infinite mixing ratios between outdoor and return air.

When outdoor temperatures sit at 52°F (just above the traditional 50°F cutoff), the system might operate at 80% outdoor air rather than switching entirely to mechanical cooling.

This staging approach extends economization by 800-1,200 hours annually compared to two-stage control.

Humidity management presents the most complex control challenge.

Direct airside economization in cold climates brings extremely dry air into facilities-often 10-15% relative humidity during winter months.

Electronics can handle low humidity, but rapid changes create static discharge risks.

Facilities address this through humidification systems, but adding moisture while simultaneously using outdoor air for cooling seems counterintuitive.

The economics still work: adding 0.03 kWh for humidification while eliminating 0.80 kWh of mechanical cooling delivers net savings, but operators must carefully track water consumption and avoid over-humidification.

Calculating PUE Improvements and ROI Quantifying economizer benefits requires understanding your facility's current energy profile and modeling performance under various outdoor conditions.

PUE (Power Usage Effectiveness) provides the standard metric, calculated as total facility power divided by IT equipment power.

A facility running at PUE 1.60 consumes 0.60 kW of infrastructure power for every 1.0 kW of IT load.

Economization reduces this infrastructure burden.

Here's a framework for calculating potential improvements for a 10 MW IT load facility in Chicago (climate zone 4A): Current State (no economization):

  • Annual IT energy: 10 MW × 8,760 hours = 87,600 MWh
  • Current PUE: 1.65
  • Total facility energy: 87,600 × 1.65 = 144,540 MWh
  • Infrastructure energy: 144,540
  • 87,600 = 56,940 MWh
  • Cooling portion (38% of infrastructure): 21,637 MWh With Indirect Airside Economization:
  • Free cooling hours: 3,400 (based on 55°F changeover)
  • Cooling energy during free cooling: 3,400 hours × 10 MW × 0.08 PUE = 2,720 MWh
  • Cooling energy without free cooling: 5,360 hours × 10 MW × 0.40 PUE = 21,440 MWh
  • Total cooling energy: 24,160 MWh
  • Cooling energy saved: 21,637
  • 24,160 = Actually increased due to fan power That calculation reveals a common error.

Free cooling isn't free-it requires fan energy to move large volumes of air.

The actual calculation must account for displaced chiller energy versus added fan energy: Corrected Calculation:

  • Baseline chiller energy: 21,637 MWh
  • Economizer mode (3,400 hours): 3,400 × 10 MW × 0.06 (fan PUE) = 2,040 MWh
  • Mechanical cooling (5,360 hours): 5,360 × 10 MW × 0.40 = 21,440 MWh
  • Total cooling energy: 23,480 MWh
  • Energy savings: -1,843 MWh (not a savings) This demonstrates why waterside economization dominates in moderate climates.

Running the same calculation with waterside economization:

  • Free cooling hours: 4,200 (wet-bulb based)
  • Economizer mode: 4,200 × 10 MW × 0.12 (pumps + towers) = 5,040 MWh
  • Mechanical cooling: 4,560 × 10 MW × 0.40 = 18,240 MWh
  • Total cooling energy: 23,280 MWh
  • **Energy savings: 21,637
  • 23,280 = Still negative** The confusion here stems from the baseline PUE distribution.

At 1.65 total PUE, cooling represents approximately 0.40 of the infrastructure overhead, but this already includes some existing economization in modern facilities.

For true baseline comparison, we need pre-economizer PUE of approximately 1.80-1.85: Accurate Baseline (PUE 1.80, no economization):

  • Infrastructure PUE: 0.80
  • Cooling portion: 0.45 (56% of infrastructure)
  • Annual cooling energy: 87,600 MWh × 0.45 = 39,420 MWh With Waterside Economization:
  • Free cooling: 4,200 × 10 MW × 0.12 = 5,040 MWh
  • Mechanical: 4,560 × 10 MW × 0.40 = 18,240 MWh
  • Total: 23,280 MWh
  • Savings: 16,140 MWh (41% reduction)
  • **New PUE: 1.80
  • 0.18 = 1.62** At $0.08/kWh average industrial electricity rate, this saves $1.29 million annually.

Capital cost for waterside economizer retrofit runs approximately $4 million ($400/kW × 10 MW), delivering a 3.1-year simple payback.

Practical Examples Example 1: Digital Realty's Chicago Campus Retrofit Digital Realty's CH1 facility opened in 2009 with traditional chilled water systems running at constant 45°F supply temperatures, achieving PUE of 1.72.

In 2019, the company invested $14 million to retrofit the 37 MW facility with waterside economization including variable speed drives on cooling tower fans, upgraded controls allowing 56-68°F supply water temperatures, and new plate-and-frame heat exchangers optimized for higher temperature differentials.

Post-retrofit data from 2020-2022 shows:

  • Average annual PUE: 1.48 (14% improvement)
  • Free cooling hours: 3,850 (44% of year)
  • Annual energy savings: 18,400 MWh
  • Cost savings: $1.47 million annually at $0.08/kWh
  • Simple payback: 9.5 years
  • Water consumption: reduced 28% through reduced cooling tower operation The project faced unexpected challenges during the first year.

Original control programming created hunting behavior-rapidly cycling between economizer and chiller modes when outdoor wet-bulb temperatures hovered near the 48°F setpoint.

This cycling stressed equipment and actually increased energy consumption during shoulder seasons.

Digital Realty implemented 15-minute delay timers and 3°F deadbands (economizer engages at 48°F but doesn't disengage until 51°F), resolving the issue and adding an additional 600 hours of economizer operation. Example 2: AWS Dublin Region Climate Optimization When AWS selected Dublin, Ireland for its eu-west-1 region expansion in 2018, free cooling potential drove the decision.

Dublin delivers 5,200+ hours annually below 55°F dry-bulb temperature, but more importantly maintains relatively consistent temperatures year-round (average high 61°F in summer, 47°F in winter).

This consistency reduces mechanical cooling system sizing requirements-AWS installed chillers sized for only 60% of peak thermal load, with economizers handling the remaining 40%.

The facility uses indirect airside economization with evaporative assist during the limited warm periods.

Operational data from 2020 showed:

  • Annual PUE: 1.21
  • Free cooling percentage: 87% of hours
  • Chiller runtime: 1,100 hours (12% of year)
  • Water consumption: 0.19 L/kWh (85% below industry average) Key to this performance is AWS's integrated approach.

Server inlet temperatures float between 68-80°F based on outdoor conditions rather than maintaining fixed setpoints.

Compute workloads route to availability zones based on current cooling efficiency-batch processing jobs preferentially run in zones experiencing optimal economization conditions.

This geographical load balancing adds approximately 400 hours of economizer operation across the region annually. Example 3: Switch TAHOE Reno Adiabatic Cooling Switch's 1.3 million square foot TAHOE Reno facility targets PUE below 1.20 using 100% outside air cooling with zero mechanical refrigeration.

The facility operates in climate zone 5B with 5,100 expected free cooling hours, but achieves economization 98% of the year through aggressive adiabatic (evaporative) cooling during summer peaks.

When outdoor temperatures exceed 75°F (roughly 300 hours annually), the facility uses direct evaporative cooling-spraying fine water mist into incoming air streams, dropping temperatures by 15-20°F through evaporation.

The physics are straightforward: converting liquid water to vapor consumes 2,260 kJ/kg, pulling heat from surrounding air.

At 95°F outdoor temperature with 15% relative humidity (typical Reno summer afternoon), evaporative cooling delivers 68-70°F supply air-adequate for maintaining data hall temperatures below 80°F.

Annual performance metrics:

  • Actual PUE: 1.18
  • Water consumption: 45 million gallons (0.23 L/kWh for 50 MW facility)
  • Mechanical cooling hours: 0
  • Air filtration costs: $380,000 annually (2.8× traditional facilities)
  • Humidity control incidents: 12 annually (requiring temporary load reductions) The facility's vulnerability emerged during the 2021 California wildfire season.

Smoke particulates overwhelmed filtration systems, forcing operators to temporarily close outdoor air dampers and operate on recirculated air with supplemental cooling.

This 72-hour event pushed monthly PUE to 1.44, demonstrating the operational risk inherent in direct economization approaches.

Common Misconceptions "Free cooling is always better than mechanical cooling" This oversimplification ignores fan power, humidity control costs, and filtration requirements.

Moving 2 million CFM of air through a facility requires substantial fan energy-typically 0.06-0.12 on the PUE scale.

In climates with limited economizer hours (below 2,000 annually), the capital cost and fan energy can exceed mechanical cooling savings.

Verizon's Miami facility evaluated airside economization in 2017 and determined the 900 annual free cooling hours would deliver only $180,000 in savings against $3.2 million capital investment-a 17+ year payback that didn't justify the project.

The facility instead invested in high-efficiency chillers with variable speed drives, achieving better ROI.

Additionally, economization requires IT equipment rated for wider temperature and humidity ranges.

Legacy equipment designed for 68-72°F, 45-55% RH environments cannot tolerate the 65-85°F swings common in economizer operations.

Facebook's early Prineville deployment faced higher-than-expected hardware failure rates during the first 18 months-approximately 2.8% annually versus the projected 2.2%-due to thermal cycling stress.

The company addressed this through modified BIOS settings that reduced CPU boost frequencies during high-temperature periods, accepting small performance reductions (3-5%) to protect hardware longevity. "Waterside economization requires cold climates" Wet-bulb temperature determines waterside economizer performance, not dry-bulb temperature.

Coastal facilities in moderate climates often outperform inland locations with colder winters.

San Francisco maintains wet-bulb temperatures below 50°F for 6,400 hours annually despite relatively mild winters (average January temperature 52°F).

The consistent marine influence keeps humidity low, enabling aggressive evaporative cooling.

CoreSite's LA1 facility in downtown Los Angeles achieves 3,800 waterside economization hours annually-more than many Chicago facilities-through ocean-influenced weather patterns.

Coastal breezes deliver afternoon wet-bulb temperatures 12-15°F below dry-bulb readings, extending economizer operation well into afternoon hours.

The facility operates chillers primarily during September-October when Southern California experiences its highest wet-bulb temperatures, not during winter months as operators might expect.

Summary & Key Takeaways

  • Airside economization uses outdoor air directly (direct) or through heat exchangers (indirect) to cool facilities, achieving PUE improvements of 0.15-0.35 in suitable climates but requiring careful contamination management and air quality monitoring
  • Waterside economization cools facility water through evaporative towers or dry coolers based on wet-bulb temperature, delivering more consistent performance across climate zones and better contamination protection at higher capital costs ($300-400/kW versus $150-200/kW for airside)
  • Climate zones 4-6 deliver optimal economization value with 3,400-5,800 annual free cooling hours, driving hyperscaler site selection toward locations like Oregon, Ireland, and Scandinavia where facilities achieve 50-84% economization time
  • Actual PUE improvements require careful calculation accounting for fan and pump power, with typical savings ranging from 0.18-0.28 PUE points (converting 1.80 baseline to 1.52-1.62) rather than eliminating all cooling overhead
  • Control strategy sophistication directly impacts economization hours gained-dynamic setpoints, variable damper positioning, and predictive algorithms add 15-25% more economizer operation compared to fixed two-stage controls
  • ROI varies dramatically by implementation approach and climate, with waterside retrofits typically delivering 3-7 year paybacks in moderate climates while airside direct economization in cold climates can achieve sub-2-year returns

Next Steps Continue building your cooling system knowledge by examining specific heat rejection technologies in the "Cooling Towers and Dry Coolers" lesson, which details the equipment that makes waterside economization possible.

The "Liquid Cooling Systems" lesson explores how direct-to-chip cooling changes economization calculations for high-density deployments exceeding 30 kW per rack.

For deeper understanding of the metrics driving these decisions, review the "PUE and Energy Efficiency Metrics" lesson for standardized measurement approaches and industry benchmarking data.