<\/p>
Until very recently, the discussion about the costs of artificial intelligence focused on GPU prices and the availability of skilled talent. Now the focus is clearly shifting toward infrastructure, which will cost less not so much up front as during use. According to the latest estimates from the International Energy Agency (IEA)<\/a><\/mark>, global energy consumption by data centers has already exceeded 460 TWh, matching the energy demand of large European economies. And that simply costs money.<\/p> In this context, the phenomenon known as “Heat Economy” is gaining prominence. Data centers are no longer viewed solely as consumers of electricity. Now, they are beginning to serve as suppliers to the district heating system. For Poland, which has one of the most extensive district heating networks on the continent, this trend opens up unprecedented opportunities for synergy between the digital and municipal sectors.<\/p> For years, data center efficiency was measured by the PUE (Power Usage Effectiveness) metric, aiming to minimize the energy used for cooling. Today, with the proliferation of liquid cooling, the industry is adopting a new standard: ERE (Energy Reuse Effectiveness). Traditional air cooling is giving way to hydronic systems in which the water removing heat from processors reaches temperatures of about 50\u201360 degrees Celsius. From a thermodynamic standpoint, this is an ideal heat source for heat pumps supplying district heating networks. As a result, computing infrastructure becomes a low-emission combined heat and power plant.<\/p> The practicality and scalability of this model can be observed in a project being carried out in Espoo, Finland, by Microsoft and Fortum. This investment, currently the largest initiative of its kind in the world, meets the heating needs of nearly 40% of the metropolitan area\u2019s residents, while reducing CO2 emissions by 400,000 tons per year<\/a><\/mark>. Similar solutions, albeit on a smaller scale, are also in place in Dublin<\/a><\/mark>, where an Amazon Web Services (AWS) server farm heats public buildings in the Tallaght district.<\/p> However, this transformation is not only driven by economic considerations but also by regulatory pressure. A key driver has been the implementation of the revised Energy Efficiency Directive (EED). According to the latest EU regulations<\/a><\/mark>, newly built data centers with a capacity exceeding 1 MW are required to conduct a rigorous cost-benefit analysis regarding the use of waste heat.<\/p> For investors, this means the end of the voluntary era. The location of new digital investments can no longer be dictated solely by access to fiber and cheap electricity. The proximity of district heating nodes, enabling the resale of thermal energy, is becoming an equally important factor. In practice, this provision compels the integration of spatial planning for IT infrastructure with municipal energy plans.<\/p> In the European context, Poland is in a unique position. Our country has one of the most developed district heating systems in the European Union. According to data from the Energy Regulatory Office (URE), we have a network of over 22,000 kilometers<\/a><\/mark>. This infrastructure, still largely powered by coal, is desperately seeking sources of decarbonization to avoid the rising costs of emission allowances (ETS).<\/p> Consequently, cooperation between data centers and district heating companies seems a natural direction for development. According to estimations, a 50 MW data center can meet the heating needs of around 15,000-20,000 households. Domestic operators are already recognizing this potential. One example is Pozna\u0144<\/a><\/mark>, where heat recovered from Beyond.pl’s data center is fed into the municipal network.<\/p> Integration with district heating systems fundamentally changes the business model for operating cloud computing. Revenues from heat sales\u2014or the avoided costs of its disposal\u2014can reduce a data center’s operating expenses (OPEX) by 15-20%, according to various estimates.<\/p> Given the costs of training AI models run into tens of millions of dollars, such optimization can determine competitiveness. Moreover, “green” heat is becoming an asset in ESG reporting, which is no small matter for global corporations.<\/p>An end to waste?<\/h4>
Brussels turns off the tap<\/h4>
Poland’s opportunity in the pipelines<\/h4>
The new digital economy<\/h4>