By the end of the decade, the carbon footprint of U.S. data centers will equal roughly 40% of all carbon emitted in the country annually, a new report from Morgan Stanley predicts.

The report projects the industry will produce 2.5 billion metric tons of greenhouse gases by 2030. That is three times more than what would have been emitted without the emergence of generative artificial intelligence — as Big Tech firms are investing billions into new data center construction — according to Reuters and The Register.

Morgan Stanley’s model looks at emissions stemming from both the construction of new data centers and the generation of the massive amounts of electricity they use. Energy use accounts for around 60% of the report’s predicted emissions totals, with the remainder stemming from the embodied carbon produced when building new data centers or manufacturing the materials and equipment for them.

The report anticipates global data center emissions rising steadily between now and the end of the decade, from 200 million metric tons this year to 600 million tons in 2030.

The advent of AI has accelerated what was already an unprecedented data center building boom. AI accounted for around 20% of new data center demand over the last year, according to JLL, with growing investments in these technologies building on the roughly $300B in hyperscale AI capital expenditures deployed in recent years.

As a result, the data center industry’s energy use, and thus its carbon emissions, also continues to skyrocket. Data centers added the equivalent of a second New York City to U.S. power grids last year alone. While data centers accounted for just 1.8% of the country’s energy consumption in 2018, that figure is expected to leap to 6.6% by 2028, according to TD Cowen.

This has created a predicament for firms like Amazon, Microsoft, Google and Meta that have driven the bulk of global AI infrastructure spending and power consumption even as they pursue ambitious and well-publicized emissions reduction goals. These companies have instead seen a jump in carbon emissions tied to data centers in the nearly two years since the start of the generative AI boom.

According to Morgan Stanley, this dynamic is likely to drive a vast increase in the market for carbon capture solutions. With the four largest hyperscalers requiring 20 times their current investment in carbon capture to offset their anticipated AI emissions, the report indicates the amount of investment required to meet these needs could reach $45B.