AI Data Centre Boom ‘Driving Huge Rise In Emissions’
AI, cloud data centre expansion to produce 2.5 billion metric tonnes of emissions through end of decade, finds Morgan Stanley report
The current expansion of data centres due to corporate demand for AI and cloud infrastructure is likely to produce around 2.5 billion metric tonnes of carbon dioxide-equivalent emissions worldwide through the end of the decade, Morgan Stanley research has found.
The emissions are likely in turn to spur investment in decarbonisation initiatives, the investment bank said in a report.
The emissions are being driven by large-scale data centre-intensive tech firms such as Google, Microsoft, Meta and Amazon, the report said.
The same companies are also holding to pledges to cut emissions by 2030, creating a “large market for decarbonisation solutions”, the report found.
Emissions surge
The emissions from the data centre industry are likely to amount to about 40 percent of what the entire US emits in a year, the report said.
The bank predicted increased demand for clean power development, energy-efficient equipment and green building materials, as well as carbon capture, utilisation, and sequestration (CCUS) technologies and carbon dioxide removal (CDR) processes.
Microsoft said in its annual sustainability report in May that its total carbon emissions had risen nearly 30 percent since 2020, primarily due to the construction of data centres.
The firm, which is one of the biggest investors in generative AI via its interests in ChatGPT developer OpenAI, said its emissions increase was largely due to building materials and hardware such as chips, servers and racks used in constructing data centres.
“Our challenges are in part unique to our position as a leading cloud supplier that is expanding its datacenters,” the company said at the time.
Green demand
It said the situation illustrated the need for greener concrete, steel, fuels and semiconductors.
Microsoft has pleged to become carbon negative by 2030.
The power consumption of Nvidia’s AI accelerator GPUs, a key component in AI data centres, jumped from its A100 to H100 chips in the space of two years, and Nvidia has indicated its next-generation Blackwell chips, announced in March, will consume even more power as they ramp up processing capacity.
At the same time, AI could reduce the demand for oil and reduce oil prices, over the next decade by improving supply and reducing costs through improved logistics and an increase in profitably recovered resources, Goldman Sachs said this week.
Goldman said it had observed a 25 percent productivity gain in early adopters of the tech in the oil and gas sector.
