According to the paper, the public health cost from data centers is 1/3 of the road transportation emissions of California. The tool they used (https://www.epa.gov/cobra) to calculate the emissions impact doesn’t have aviation emissions.
Following from your model California is responsible for roughly 9.6% of the total road transportation emissions in the U.S > So 1/3rd of that is 3.2% . So road transportation's health cost is roughly 30x higher than datacenters'. Aviation, while extremely polluting (~9%), is also dwarfed by road transport due to the latter's shear volume
Another thing to observe: The paper's claims are based on a McKinsey projection to 2030 that assumes 'a strong decrease of road transport emissions' and a 'surging demand for AI data centers' that 'outweighs power plants emission efficiency improvement'.
Contrast this to today where energy estimates are 2% for datacenters vs. 80% for road transport. Guess what is causing more outweighing of power plant emission efficiency improvement as that fleet of cars, trucks and lorries is electrifying (the reason for McKinsey's reduction projection)?
Now I'm not saying we should let data-centers get a free pass. Pollution is a serious problem. But in terms of priorities, curbing datacenters might not be the number one concern.
