数据中心冷水机组：“自由冷却”对冷水机组服务收入的影响有多大

Data Center Chillers (3/3): How much can "free cooling" impactchiller service revenues? In our previous articles in this series on data center cooling, we have covered howchillers work, the difference between air-cooled and water-cooled units (along with theireconomics), and run through “chiller gate” risk on equipment sales from both climatic andoperational risk perspectives. However, equipment sales are only one part of chiller gate.The other question we often get asked is the impact reduced compressor use could have on Varun Govindaraj+1 917 344 8543varun.govindaraj@bernsteinsg.com Steve Song+1 917 344 8401steve.song@bernsteinsg.com Before we can understand which parts of chiller service revenue are at risk, we first needto understand its constituent parts. There are four broad buckets. Not every installed chillergenerates all four of these revenue streams, but we can assume representative averages. First, there is maintenance revenue- largely preventative and delivered through plannedservice agreements where technicians visit a unit at regular intervals. This could bemanaged through an up front payment or annual contracts and the duration can also varybefore renewal. Next you haverevenue from equipment modernization / upgrades;these only happen once or twice over the life of a chiller (~25 years) and take place whena piece of technology is replaced (e.g., sensors added, compressor is upgraded). The lastmajor bucket isrepair revenue from equipment operational issues or breakdowns.Repair revenue tends to increase in share as the unit gets older. In some cases theremay We believe maintenance and modernization are not at risk from reduced usage of thechiller, while repair (both labor and parts) is and amounts to somewhere between 25 - 30%of the total revenue pool.This does not mean the entire repair revenue pool itself is at risk of going away, only that it will likely see some headwinds from reduced usage.This is also not likley to be an immediate headwind, the chiller needsto run for 5 - 10 years before there’s enough wear on the compressor to increase likelihood of breakdown.We estimatea ~1 - 2% headwind to the DC service revenue toplineafter that. Howmuch depends on the extent of free cooling that we see ; for example, chillers installed inNoVa will probably break down a lot less than those installed in Phoenix. But given we are (discussing how chillers work, and the different economics of air cooled and water cooledunits) and part two (discussing the risk of chiller gate on equipment sales through climatic BERNSTEIN TICKER TABLE INVESTMENT IMPLICATIONS We rate TTOutperformwith a target price of$550. We rate CARRMarket-performwith a target price of$75. We rate VRTOutperformwith a target price of$416. DETAILS EXHIBIT 2:Sizeable part of chiller service revenues could be at risk from free-cooling In our previous articles in this series on data center cooling, we have covered how chillers work, the difference betweenair-cooled and water-cooled units (along with their economics), and run through “chiller gate” risk from both climatic andoperational risk perspectives. However, equipment revenues are only one part of chiller gate. The other side we often get asked What is free-cooling? First, let’s clarify what we mean by free-cooling. In certain situations, the ambient temperature at the location a data center hasbeen built can drop to a point where the chiller does not need to run at full capacity (or run at all) for cooling outcomes to beachieved. This can happen in one of three ways. The first is through air-side economizers, where cold air from outside the datacenter is directly piped in to cool the room. While this works with lower rack densities, for current Blackwell and higher unitsit is not sufficient. Then there are water-side economizers, where either cold ambient air or cold water from a cooling towercan cool the facility water system (FWS) loop directly; this also reduces or eliminates the need for a chiller to make use of its The reason this is relevant is that the compressor is the most energy intensive and load-bearing part of a chiller; if it is used lessintensively, it sees less wear and has a lower probability of breakdown. This could impact repair revenue for the OEM. And given8 - 10x the capex of a chiller is realized as nominal (not inflation adjusted) service revenue, it could be a material number. Components of service revenue Before we can understand which parts of service revenue is at risk, let’s understand how to break service revenues up. Thereare four broad buckets. Not every installed chiller generates all four of these revenue streams, but we can assume averageshere. First, there’s maintenance revenue - largely preventative and delivered through planned service agreements wheretechnicians visit a unit at regular intervals. This could be managed through an up front payment or annual contracts and theduration can also vary before renewal. Next you have reven