Datacenters to emit 3x more carbon dioxide because of generative AI
Carbon capture outfits could be making up to $45B, say analysts
The datacenter industry is set to emit 2.5 billion tonnes of greenhouse gas (GHG) emissions worldwide between now and the end of the decade, three times more than if generative AI had not been developed.
This latest contribution to the debate on bit barns and their effect on the environment comes from investment bank and financial services biz Morgan Stanley, which has developed its own model of the ecosystem and concluded there is a large market for decarbonization solutions - and money to be made - as operators strive to meet their 2030 carbon neutrality goals.
The research, "Global Data Centers: Sizing & Solving for CO2," shared privately, estimates the total global emissions for datacenters across Scope 1, 2 & 3 between now and the end of the decade, and concludes that construction of new facilities combined with their electricity needs will hit 2.5 billion tonnes of CO2-equivalent, or about 40 percent of the entire current annual emissions from the United States.
Looking across Scope 1, 2 and 3 is vital, as can be seen from Microsoft's sustainability report earlier this year. This found that Microsoft's CO2 emissions rocketed 29.1 percent from the 2020 baseline, but much of this was due to indirect emissions (Scope 3) from the construction of more datacenters to meet customer demand for cloud services.
Morgan Stanley forecasts that global CO2 emissions from bit barn construction will rise from about 200 million tonnes this year to about 600 million tonnes by 2030, a figure it claims comes out three times higher when compared to a scenario in which no construction due to extra demand from GenAI happens.
Looking at that total 2.5 billion tonnes of CO2, the forecast estimates that about 60 percent of this can be attributed to the power requirements of the facilities as operational carbon, while the remainder is due to embodied carbon, emitted during the manufacturing of construction materials and the infrastructure for those bit barns.
Much of the Morgan Stanley report is unsurprisingly aimed at investors and focuses on the opportunities for decarbonization solutions including carbon capture, utilization, and sequestration (CCUS); and carbon dioxide removal (CDR) in the datacenter market.
The banker estimates a theoretical market for CCUS by 2030 of about 260 megatonnes of CO2-equivalent for bit barn power decarbonization alone, with a $15-45 billion investment need. More than 50 percent of the global datacenter build-out is expected to be happen in the US, and Morgan Stanley says this market is particularly well suited to CCUS.
The demand for carbon dioxide removal (CDR) for just four of the hyperscale operators (Microsoft, Google, Meta and Apple) is estimated at a combined total of 25 megatonnes CO2-equivalent by 2030, and Morgan Stanley says this is about 20 times what they obtained in their most recent fiscal year.
- Don't blame AI for rise in carbon emissions, says Google exec
- Google's Irish bit barn plans denied over eco shortfall
- Rising costs biggest issue for datacenter operators as demand grows
- Small datacenters face the axe under China's new energy policy
In light of the limited supply of CDRs and the estimated growth in datacenter emissions, Morgan Stanley expects reforestation projects to be key beneficiaries of the 2030 net-zero targets that hyperscalers have committed to.
Demand for space in bit barns is currently at a high in many markets around the globe because of the AI boom, as The Register has previously reported, with hyperscalers snapping up much of the available capacity.
In fact, the total capacity of hyperscale facilities is set to grow almost threefold over the next six years on the back of AI demand, substantially increasing the amount of power required by those sites and sparking development of new ones. ®