Global data centers to raise CO2 emissions dramatically

Morgan Stanley looks at the cumulative global carbon footprint through 2030, and sees this creating a huge market for decarbonization solutions as many large cloud computing companies look to achieve carbon neutrality goals.

The US bank said in a note dated September 2 that the construction of global data centers through 2030, in addition to their electricity needs, will offset more than 40% of the United States’ greenhouse gas emissions in one year, or nearly 2.5 billion tons of carbon dioxide equivalent.

“We expect global greenhouse gas emissions from developing countries to rise from around 200 million tonnes of carbon dioxide equivalent in 2024 to around 600 million tonnes of carbon dioxide equivalent in 2030 – representing emissions three times higher in 2030 than in a scenario without AI,” the bank said.

“We believe that the volume of greenhouse gas emissions is larger than estimated, as well as the share of embodied carbon at around 40%, and will drive significant growth opportunities for DC ‘decarbonization solutions’ business models,” the bank said, referring to clean energy, energy-efficient equipment, green materials, carbon capture, utilization and storage, and carbon dioxide removal solutions.

The bank sees reforestation projects as a major beneficiary of ambitious 2030 net-zero emissions targets set by corporate giants due to the limited supply of technology-based decarbonization.

Governments are unlikely to seek to limit data center development given the many benefits associated with pursuing AI, including broad economic efficiencies, drug discovery, potential long-term decarbonization benefits through efficiencies in complex systems such as cities and transportation, impacts on behavior change through providing more sustainable solutions to customers, and the discovery of new technologies, among others.

“This underscores the importance of DC decarbonization solutions,” Morgan Stanley added.

centersCO2DatadramaticallyemissionsGlobalRaise