* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

For Q3, our fleet-wide TTM PUE was 1.12, which has remained constant for the past eight quarters. Our fleet-wide quarterly PUE was 1.13, which is down from 1.14 in Q3 of last year.

Our lowest TTM PUE was 1.10, achieved by five facilities. Our lowest quarterly PUE for an individual facility was 1.09, achieved by Data Centers M and N.

Our highest TTM PUE for an individual facility was 1.13, at Data Centers C, D, G, H. Our highest quarterly PUE for an individual facility was 1.26, for Data Center U. This facility is not currently supporting a full IT load, and we expect the PUE to improve when the facility runs closer to its design load.

This quarter, we removed Data Center B from our public reporting, since it is being decommissioned. We also added new facilities to our public reporting.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

For Q2, our fleet-wide quarterly PUE was 1.11, and our fleet-wide TTM PUE was 1.12. These are the same as the previous quarter's results.

In Q2, we had five sites reporting quarterly PUEs of 1.10 or less. Three sites had TTM PUEs of 1.10 or less in the same period.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

For Q1, our quarterly PUE across the fleet was flat from the prior quarter, and remained at 1.11. The corresponding fleet-wide TTM PUE was 1.12, which matches what it was a year ago.

New this quarter is that we have removed Data Center A from reporting. For the past year we have shared that we were moving load out of this site, and in Q1 the load dropped below our 5 MW threshold for reporting. As a result of this change, our maximum site quarterly PUE has dropped from 1.21, which was data center A in Q4 2013, to 1.16, which was data center F in Q1 2014.

In Q1, we had six data centers reporting quarterly PUE of 1.10 or less, with five coming in at 1.09.

For TTM site PUE, data centers J, L and N repeated their Q4 performance by producing the lowest TTM PUEs in our fleet, at 1.10.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

We periodically review our PUE calculations to ensure consistency with The Green Grid guidelines for each site and consistency across sites. In Q4 2013, we updated our PUE formulas for 2013 to include more of our site infrastructure as overhead, which has resulted in many of our PUE values increasing slightly. In addition to mechanical and electrical losses, PUE overhead energy in our PUE calculations includes energy used in substations, transformers, water treatment plants, accessory buildings (such as employee cafeterias and gyms) and office space. Yes, that's more than we've seen anyone include in their PUE. We could strip all this out and have a PUE of 1.06 or less, but we believe it's better to measure--and optimize--all aspects of data center operations. All 2013 values have been updated to reflect the new calculations.

For Q4 2013, our fleet-wide quarterly PUE dropped to 1.11, and fleet-wide TTM remained flat at 1.12.

On an individual site basis, two sites had PUE of 1.09, and seven had PUE of 1.10 or less. Data center A continued to have the highest PUE at 1.21.

For our TTM PUEs, three sites had TTM PUEs of 1.10 or less.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

For Q3, our quarterly PUE across the fleet went up by 0.02, which was expected since Q3 has the hottest months of the year. The corresponding fleet-wide TTM PUE was 1.12, the same as the prior quarter.

As for the individual sites themselves, all but one of our sites continued to report quarterly PUE less than 1.20. Two of our sites have PUE of 1.10 or lower.

For individual site TTM PUE, three data centers reported a TTM PUE of 1.10 or lower.

We periodically review our PUE calculations to ensure consistency with The Green Grid guidelines for each site and consistency across sites. In Q4 2013, we updated our PUE formulas for 2013 to include more of our site infrastructure as overhead, which has resulted in many of our PUE values increasing slightly. In addition to mechanical and electrical losses, PUE overhead energy in our PUE calculations includes energy used in substations, transformers, water treatment plants, accessory buildings (such as employee cafeterias and gyms) and office space. Yes, that's more than we've seen anyone include in their PUE. We could strip all this out and have a PUE of 1.06 or less, but we believe it's better to measure--and optimize--all aspects of data center operations. All 2013 values have been updated to reflect the new calculations.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

For Q2, our quarterly PUE across the fleet remained flat at 1.12, as did our fleet TTM PUE.

On an individual site basis, three of our sites had PUE of 1.10 or less. Only data center A reported a PUE greater than 1.20.

For our individual site TTM PUE, three sites had 12-month TTM PUE of 1.10 or less. Only data center A reported a TTM PUE greater than 1.20.

We periodically review our PUE calculations to ensure consistency with The Green Grid guidelines for each site and consistency across sites. In Q4 2013, we updated our PUE formulas for 2013 to include more of our site infrastructure as overhead, which has resulted in many of our PUE values increasing slightly. In addition to mechanical and electrical losses, PUE overhead energy in our PUE calculations includes energy used in substations, transformers, water treatment plants, accessory buildings (such as employee cafeterias and gyms) and office space. Yes, that's more than we've seen anyone include in their PUE. We could strip all this out and have a PUE of 1.06 or less, but we believe it's better to measure--and optimize--all aspects of data center operations. All 2013 values have been updated to reflect the new calculations.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

For Q1, our fleet Quarterly PUE as well as fleet TTM PUE was 1.12. Compared to Q4 2012, our quarterly PUE is up by .01, and TTM remained flat.

As for the individual sites themselves, six of our sites posted a Quarterly PUE of 1.10 or less. All but one of our sites continued to report quarterly PUE less than 1.20.

For individual facility TTM PUE, all but one of our facilities showed PUE improvement as compared to Q1 2012. Data Center J reported a TTM PUE of 1.09, which is the same as its Q4 2012 TTM. Data center A continued to run at our highest PUE of 1.22.

We periodically review our PUE calculations to ensure consistency with The Green Grid guidelines for each site and consistency across sites. In Q4 2013, we updated our PUE formulas for 2013 to include more of our site infrastructure as overhead, which has resulted in many of our PUE values increasing slightly. In addition to mechanical and electrical losses, PUE overhead energy in our PUE calculations includes energy used in substations, transformers, water treatment plants, accessory buildings (such as employee cafeterias and gyms) and office space. Yes, that's more than we've seen anyone include in their PUE. We could strip all this out and have a PUE of 1.06 or less, but we believe it's better to measure--and optimize--all aspects of data center operations. All 2013 values have been updated to reflect the new calculations.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

In Q4, our energy-weighted quarterly PUE dropped .03 to 1.11 from a Q3 value of 1.14. This ties with Q1 2012 as our best quarterly energy-weighted PUE value so far. Each of our sites had quarterly PUE values less than 1.20, with five sites reporting quarterly PUE values less than 1.10. For the first time, two of our sites reported PUE values of 1.08.

Our individual facility maximum quarterly PUE dropped from 1.25 to 1.19. This drop resulted from our work at Data Center A, where we had been moving IT load out to more efficient facilities over the past few quarters. The PUE at that site initially went up, but once we had moved enough IT load out of that site, we were able to reduce our cooling to better match the remaining IT load — resulting in an overall drop in PUE.

Our TTM energy-weighted PUE for Q4 was our best on record at 1.12, which was down .01 from last quarter. All but one of our sites reduced their TTM PUE values in Q4 2012, compared to Q4 2011. Data center J reported a TTM PUE of 1.09, making it our first site to report a TTM PUE less than 1.10.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

In Q3 2012, our fleet-wide quarterly PUE increased by 0.02 from 1.12 to 1.14. We expect to see a slight increase in PUE as the weather warms between Q2 and Q3 in the Northern Hemisphere, and the 0.02 Q2-Q3 increase this year is in line with prior year increases. As compared to Q3 2011, all but one of our sites had better quarterly PUE results than they did a year ago.

Fleet-wide TTM PUE performance remained flat at 1.13 from Q2 to Q3.

The quarterly PUE at data center J dropped from 1.09 to 1.08 in Q3, which makes it our best performing data center. The TTM PUE for that same data center was 1.10, which was also our best TTM PUE. Ongoing reconfigurations at data center A resulted in our highest quarterly PUE of 1.25.

Note: We caught some miscalculations and have made corrections to our Q3 2012 PUE results since we first posted them in October 2012. We updated our quarterly PUE from 1.13 to 1.14, our TTM avg. PUE from 1.12 to 1.13, our individual facility min. TTM PUE from 1.08 to 1.10, our individual facility max. quarterly PUE from 1.18 to 1.25, and our individual facility max. TTM PUE from 1.18 to 1.23.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

In Q2 2012, our quarterly fleet PUE increased by 0.01 from 1.12 to 1.13. We expect to see a slight increase in PUE as the weather warms between Q1 and Q2 in the Northern Hemisphere, and this year the 0.01 Q1-Q2 increase was the smallest increase we've seen. Individually, eight of ten sites had better quarterly PUE results than they did in Q2 2011.

We included our fourteenth facility, Data Center N, because it meets our PUE reporting criteria of having 6 months of operation above 5 MW of load. Data Center N posted a quarterly energy-weighted PUE of 1.12, which is lower than our quarterly fleet PUE of 1.13.

Ongoing cooling system retrofits at data centers D and E reduced quarterly PUEs from 1.20 and 1.18 last year to 1.15 in Q2 2012. Data center A reported a higher than typical quarterly PUE of 1.23 as we continue to move IT load out of that site to more efficient facilities.

Fleet TTM performance held steady at 1.13 from Q1 to Q2.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

In Q1 2012, we posted another best quarterly fleet PUE performance at 1.11, down from last quarter's fleet PUE of 1.12. Seven different facilities posted quarterly best results, and four facilities returned quarterly energy-weighted PUEs of less than 1.10.

We included our twelfth facility, Data Center M, because it meets our PUE reporting criteria. Data Center M posted a quarterly energy-weighted PUE of 1.14, higher than the steady-state design target but inline with expectations due to commissioning activities.

Cooling system retrofits at data centers D and E reduced quarterly PUEs from 1.22 and 1.19 last year to 1.14 in Q1 2012. Data Center A reported a higher than typical quarterly PUE of 1.22 due to site work.

Fleet TTM performance also set a new best as our energy-weighted average PUE improved from 1.14 in Q4 to 1.13. Q1 also marked a full year since we added natural gas and office space to our overhead — yet both quarterly and TTM PUEs continue to improve.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

We posted our best quarterly PUE performance to date at 1.12, down from the previous best of 1.13 in Q1 2010. We accomplished this even after adding two new facilities and even during a quarter with warmer than average temperatures. Data Center F also set a new quarterly best PUE performance for an individual facility at 1.08, improving on the previous best of 1.09 by Data Center E in Q1 2011.

As mentioned, this quarter we included two additional facilities in our calculations because they meet our PUE reporting criteria (more than 5MW of critical load and operating for more than 6 months). New Data Centers K and L posted quarterly energy-weighted PUEs of 1.13 and 1.15 respectively. New facilities typically result in higher-than-design PUE because of commissioning activities, partial loading, and a lack of operational tuning.

Comparing 2011's TTM performance to 2010's, our fleet energy-weighted average PUE improved from 1.16 to 1.14. For individual facilities, Data Centers E, F, and J are all operating with a TTM PUE of 1.11, while only Data Center J operated at a TTM PUE of 1.12 in 2010.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

Typical warmer summer temperatures increased our fleet's quarterly energy-weighted PUE to 1.16 from 1.14 in Q2. Our quarterly fleet performance showed a significant improvement YoY from 1.20 in Q3 2010 due to less extreme weather and ongoing operational improvements.

Weather almost invariably makes Q3 post the highest quarterly energy-weighted PUE of the year. Yet our Q3 2011 results, the worst quarterly performance of the year (assuming we don't have an abnormal Q4), is the same or better than all but our best quarterly performance in 2008, 2009, and the first half of 2010.

Our trailing twelve-month energy-weighted average PUE continues to improve with our fleet performance being 1.14, the lowest level since we began reporting PUE data. The TTM energy-weighted PUE has improved every quarter since Q3 2009, and has improved from 1.17 YoY.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

In Q2 we updated how we calculate PUE to comply with The Green Grid's revised recommendations on measuring and reporting PUE. The change added two new overhead components: natural gas and electricity consumed by data center office space. These changes will have a modest negative impact on PUE. From Q2 onward, we will use the new methodology to report our PUE. However, we will not recalculate Q1 2011 and previous quarters to comply with the new guidelines.

Even with the inclusion of additional overhead terms, the quarterly energy-weighted average PUE was our best Q2 result to date at 1.14. We saw a small increase from 1.13 in Q'11, a strong performance given warmer Q2 temperatures. Quarterly fleet performance improved YoY from 1.17, driven by continued operational improvements, more favorable weather (compared to Q2 2010), and facilities reaching design targets. All but one facility improved YoY PUE for Q2.

Our trailing twelve-month energy-weighted average PUE continues to improve with our fleet performance being 1.15, the lowest level since we began reporting PUE data. The TTM energy-weighted PUE has improved every quarter since Q3 2009, and has improved from 1.17 YoY.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

Our Q1 '11 quarterly energy-weighted average PUE of 1.13 is another excellent fleet performance, matching our best previous in Q4 2010. Typical Q1 seasonal effects kept PUEs low. Data Center E set a new best energy-weighted quarterly PUE to date of 1.09 and Data Center J set a new best TTM PUE of 1.11. Both are excellent performances that are approaching facility design targets.

Q1 '11 TTM energy-weighted average PUE for our fleet also stayed at the previous low 1.16 set last quarter. YoY only data centers C and D had increased quarterly PUEs, in part due to retrofit work being done. All other facilities posted improved PUEs YoY due to continued optimization in facility operations.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

Our Q4 '10 quarterly energy-weighted average PUE of 1.13 is the best result since we started reporting, beating the PUE of 1.15 in Q1 '10. All but one facility (Data Center C) reported sub-1.20 PUE, and half of the facilities reported PUEs of less than 1.12. Typical seasonal effects were the primary driver for QoQ improvements.

The Q4 '10 TTM energy-weighted average PUE also set a new low of 1.16, down from 1.17 in Q3, highlighting steady continuous improvement of the entire fleet. YoY, only one facility (Data Center C), currently undergoing a retrofit, reported a higher PUE in Q4'10 than Q4 '09 â€" all other facilities improved their efficiency. We achieved YoY improvements by improving execution on optimal operational practices and having our facilities operate closer to the design targets.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

On average, the Q3 wet-bulb temperatures at our facilities were the highest since we started reporting PUE. This weather drove an expected QoQ increase in the total quarterly energy-weighted average PUE to 1.20 but the TTM PUE remained flat at 1.18. Even with unfavorable weather, our fleet's quarterly performance improved YoY from 1.22 to 1.20 and the TTM performance improved from 1.19 to 1.18. Data Center J's continued operational tuning resulted in a quarterly (and facility) best PUE at 1.12, a strong performance.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

Typical seasonal effects resulted in a QoQ increase in the total quarterly energy-weighted average PUE to 1.17. On average, the Q2 wet-bulb temperatures at our facility locations were higher than those of 2009, yet our fleet's quarterly performance improved YoY from 1.20. For the first time Data Center J posted the best quarterly PUE at 1.13 due to improved operations and tuning. The trailing twelve-month energy-weighted average PUE for all facilities remained at 1.18, an improvement from 1.19 in Q2 2009. Peak summer weather will likely see an increase in the fleet's quarterly energy-weighted average PUE for Q3.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

Q1 '10 was the best quarterly fleet performance recorded, with the quarterly energy-weighted average PUE of all facilities falling to 1.15 from 1.16 in Q4. All but one facility posted a PUE of 1.2 or less. The improvement was due to the typical seasonal effects and improved operational strategies at Data Centers G, H, I, and J. Data Center E also posted the best quarterly PUE performance from an individual facility to date at 1.10. The trailing twelve-month energy-weighted average PUE improved to 1.18 from 1.19 the same quarter a year ago. We expect higher PUE values in Q2 due to less favorable ambient weather conditions.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

Seasonal effects and operational improvements enabled QoQ efficiency improvements for all data centers in Q4, reducing the quarterly energy-weighted average PUE of all facilities to 1.16 from 1.22 in Q3. The best quarterly efficiency performance from an individual facility was a PUE of 1.11 (Data Center E), tying our lowest reported value to date (Data Center B, Q1 '09). The trailing twelve-month energy-weighted average PUE improved to 1.19 from 1.20 the same quarter a year ago. The best TTM energy-weighted PUE for an individual facility dropped to 1.13 (Data Center B), the lowest TTM PUE to date. Data Centers G, H, I, and J continue to mature operationally, improving their PUE. We expect this trend to continue until they reach the steady state design targets.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

We added one new facility, Data Center J, to our PUE report. Overall, our fleet QoQ results were as expected. The Q3 total quarterly energy-weighted average PUE of 1.22 was higher than the Q2 result of 1.20 due to expected seasonal effects. The trailing twelve-month energy-weighted average PUE remained constant at 1.19. YoY performance improved from facility tuning and continued application of best practices. The quarterly energy-weighted average PUE improved from 1.23 in Q3'08, and the TTM PUE improved from 1.21. New data centers G, H, I, and J reported elevated PUE results as we continue to tune operations to meet steady-state design targets.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

This quarter three new data centers (G,H, and I) met inclusion criteria, bringing the total number of conforming facilities to nine. The PUE of the new facilities is elevated due to commissioning and bring-up activities. The combined impact of partially-commissioned facilities and traditional seasonal effects resulted in a QoQ increase of our quarterly energy-weighted average PUE, bringing it to 1.20. The TTM energy-weighted average PUE for all conforming facilities, however, remained at 1.19, despite warmer temperatures. This result is due to continued YoY operational improvements for Google's fully commissioned data centers. Both the minimum and maximum individual facility TTM PUE results decreased in this quarter as a result of operational improvements.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

The efficiency of all data centers again improved QoQ due to further optimization and cooler temperatures. The quarterly energy-weighted average PUE dropped to 1.15 from 1.16 in Q4, and the quarterly PUE values for each data center was less than or equal to 1.20. The collective TTM energy-weighted average PUE fell to 1.19 from 1.20. The best-performing individual facility quarterly PUE dropped to 1.11 (Data Center F) from 1.13 in Q4 (Data Center E). Data Center F had the highest PUE in Q3 '08 due to a low IT load and incomplete commissioning activity at that time. Data Center A yielded YoY reductions due to completed cooling plant improvements. Data Centers C and D improved YoY because of increased IT load, which better utilized the cooling plant resources.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

Efficiency of all data centers improved QoQ due to continued optimization and cooler temperatures. The quarterly energy-weighted average PUE dropped to 1.16 from 1.22 in Q3 while the collective TTM energy-weighted average PUE fell to 1.20 from 1.21. The best performing individual facility quarterly PUE dropped to 1.13 (Data Centers E, F) from 1.17 in Q3 (Data Center B). All facilities made YoY efficiency improvements, with Data Center A showing strong results from cooling plant improvements designed to improve free cooling.

Data Centers E and F moved through commissioning and into full operation, both yielding an excellent quarterly PUE score of 1.13 and 1.14, respectively. Looking back, Figure 2 shows the large daily variations and higher PUE quieting to an expected lower value.During commissioning, at times all chillers and air movers were run simultaneously in order to test proper operation of the facility. Concurrently, IT equipment load was small because the facility was in the early stages of deployment. These factors caused the higher Q3'08 average PUE values as well as significant day-to-day variations. Q4'08 saw these commissioning activities taper off, and the favorable seasonal conditions led to lower, stable PUE values.

* We report Individual Facility TTM PUE only for facilities with at least twelve months of operation.

Details:

Initial public release of PUE data reported TTM energy-weighted PUE values for six Google data centers satisfying the inclusion criteria. Data centers E & F show the effects of commissioning including large day-to-day PUE fluctuations and are running above the steady-state design efficiency. Seasonal influences can be seen in the Q2-Q3 time frame across all data centers.