//php echo do_shortcode(‘[responsivevoice_button voice=”US English Male” buttontext=”Listen to Post”]’) ?>
At first glance, the megawatts of heat given off by data centers (server farms) seem like a real waste of energy, yet they also represent a unique opportunity for recovery, harvesting and reuse—an opportunity that companies like Facebook’s parent, Meta, have taken in full stride.
Since 2020, Meta has been recovering excess heat from its data center in Odense, Denmark (there’s an interesting BBC video tour of Meta’s facility here). Similarly, there are about 10 other full-scale projects from other data center operators in Holland, as well as others primarily in Europe.
The industry giant describes the heat-recovery approach (see Figure 1) in simple terms: “The process to deliver heat to the community begins with a wind turbine. Multiple wind turbines create and add renewable energy to the electric grid powering our facility, including our servers. We will direct air heated by the servers over water coils, which recover the heat by raising the temperature of the water. A heat-pump facility powered by renewable energy raises the temperature of the water even more. The hot water is then delivered to the district heating network and distributed to the local community.”
The system recovers 100,000 MWh of energy per year, which is enough heat to warm 6,900 homes in that immediate community, according to Meta. The additional public-relations angle is also unique: Besides the fact that it will save energy, the pitch is that “every post warms a Dane,” because whenever someone sends an email, posts to a website or surfs the web, they are not wasting energy; they are helping heat homes.
But a hot-air-to-hotter-water system is complicated to implement. The heat is collected by circulating water through the data center via insulated steel pipes, which pass through copper coils located inside each of the data center’s 176 cooling units. The water picks up low-temperature heat at about 27˚C (80˚F) via a heat exchanger and channels it to a co-located heat-pump facility. There, the heat pump “amplifies” the warm water’s temperature—and thus its heat-based energy density—to about 70˚C to 75˚C (about 160˚F to 170˚F). If it sounds complicated, it is.
What makes such a system viable is that Denmark and other Scandinavian countries often already have a centralized hot water system in place for community heating, along with the associated underground piping—houses do not have to have their own area- or water-heating system. Note that many university campuses around the world also function this way, with a central hot water and steam plant supplying buildings around the campus with heat via insulated pipes and tunnels. This is less expensive than having independent heating units in each building and makes maintenance easier thanks to a centralized design.
Data centers, which are in proximity to these district heating systems, have the potential to provide as much as roughly 50 TWh a year of excess heat, according to a study from ReUseHeat, an EU-funded project aimed at promoting waste heat reuse. That would work out to between 2% and 3% of the energy that EU households used on space heating in 2020.
However attractive this data-center–harvesting process seems due to its perceived cost savings, it’s not an easy or wide-scale solution for several reasons. First, many locales do not have the community-based hot water and heating systems with pipes running underground to each house. Second, it takes a lot of distributed hardware and expense to capture and channel low-temperature heat, especially when it is spread out over a large area, such as a data center. There are a lot of moving parts in this approach, both literally and figuratively (see Figure 2).
The laws of thermal physics dictate—in a manner somewhat analogous to electrical energy and voltage/current levels—that energy capture, transfer and conversion is much more efficient at higher temperatures. While industrial facilities and even some offices have been using “co-generation” for many years to capture and reuse heat thrown off by equipment, that heat source is often highly coalized, as is the load being supplied by the recovered energy.
It would be nice if there was a way to effectively transform waste heat at relatively low temperatures directly to electricity at the source. Then it could be converted and regulated using electronic systems like those used in battery-based energy-storage systems, for example, and transmitted via power lines rather than as heated water. An all-electronic system would have few or no moving parts, whereas the water-based system needs pumps, heat exchangers and all sorts of bulky, maintenance-prone facilities.
What’s your view on the actual wider practicality of energy harvesting and recovery using waste heat from low-temperature sources for a wider set of users? Are these economically and technically attractive installations, or are they really “showcase” projects that enable favorable public relations and attention but will require considerable and costly support over the years to continue? It’s the inevitable real-world question: Is the pain worth the gain?
- Ramboll Group A/S, “World’s largest heat extraction installation for a hyperscale data centre”
- Datacenter Forum, “BBC: How Meta’s datacenter gives Denmark’s Odense heat and hot water”
- Meta, “Odense: Meta Data Centers”
- Meta, “This data center will warm their hearts, or at least heat their homes”
- Meta, “Facebook’s hyperscale data center warms Odense”
- Data Center Dynamics, “Facebook plugs its Danish data center into Odense district heating system”