188,560 passengers flying from New York to London annually: how built world AI can offset its carbon footprint (and some)
In Q1 2024, Pi Labs will publicly release our white paper Sustainably intelligent: AI for a greener built world. In it, we identify that if each employee within the global real estate sector were to use generative AI tools equivalent to ChatGPT-4 for an average of one hour each day, the processing would consume enough energy (and resultant carbon emissions) to match 188,560 passengers flying one-way between New York and London each year. But what environmental benefits are generated by the use of AI, and can its carbon footprint be offset? Overwhelmingly.
Concerns about energy consumption often accompany power-intensive innovations. This emerged for blockchain, and more recently generative AI innovations such as ChatGPT. Energy consumption by the hardware used for generative AI is relevant to the built environment for two key reasons: (i) anecdotal evidence suggests widespread adoption of generative AI by workers within the sector, meaning the excess energy consumption will increase a firm’s scope three emissions; and (ii) data centres housed within buildings are a growing area of interest and relevance to the real estate sector. Attempts have been made to locate data centres in sustainable and/or low-cost energy locations. Heat transfer innovations are also relevant to the built environment, where the waste heat generated by data centres within a district or building is utilised for space heating. This has been accompanied by decentralised data centre initiatives, whereby “micro data centres” replace space heating within homes or neighbourhoods. CogNovum’s Roelof Opperman is a bastion of knowledge in this area.
Assuming a figure of 1.12Wh for every request on a generative AI platform equivalent to ChatGPT-4, and an average request rate of one every two minutes, one hour of use by a single user would result in 0.034 kWh energy demand. According to Eurostat, there are 2,217,200 workers employed in the real estate sector within the European Union and United Kingdom. Therefore, an hour of daily use of ChatGPT-4 and equivalent generative AI platforms across each worker within the European Union and United Kingdom’s real estate sector is equivalent to 75,385 kWh of energy demand each day, or 22.46 tonnes of CO2-e emissions daily. Assuming the same ratio of population and real estate sector workers globally, 1,184,904 kWh (1.184 GWh) of energy demand would be created by the sector each day. This would be equivalent to 516.6 tonnes of CO2-e emissions daily, or 188,560 one-way flights between New York and London annually.
Net negative emissions
An average of an hour of daily AI use across the global real estate sector may add 188,559 tonnes of CO2-e to the annual inventory, but the construction and operation of buildings is producing approximately 15.8 gigatonnes of CO2-e annually as it is. So what if the productivity improvements and technological advancements created through the wider adoption of AI in the built environment create disproportionate counterweights to the sector’s total emissions and wider environmental footprint, accelerating its transition to a more sustainable state? Simply put, can an average of one hour of daily use of processing-intensive AI per person within the real estate sector result in reducing the sector’s emissions by more than 0.0014 percent?
Wiping a US-sized carbon footprint from the built environment
In Chapter 2 of Sustainably intelligent, we map current and future environmental AI use cases along the real estate value chain. We then focus on four specific use cases, calculating how AI adoption in each area could reduce the built environment’s annual carbon footprint. We found that by simply addressing these four use cases, 5.81 to 6.46 gigatonnes CO2-e of greenhouse gas (GHG) emissions can be avoided annually by 2030. This would offset the entire annual carbon footprint of the United States (6.02 gigatonnes in 2022)— the world’s second largest GHG emitter behind China.
Be among the first to receive Sustainably intelligent: AI for a greener built world direct to your inbox by registering here.
