"A newbie question
> leaving BNSF with the least well run customers
How does that impact operating BNSF earnings?"

The assumption is that if trains don't leave until all customers' freight is on board and a sufficiently long train is built in the yard (admittedly some exaggeration here), and poorly run customers frequently don't deliver their freight on time or in quantities expected or properly packaged and labeled, then the cost to serve those customers is higher and service interruptions more frequent.

"Wouldn't there be significant capex to change to PSR e.g. locomotives, huge new software packages etc?"

I'm not a RR guy and certainly no expert on PSR but I think the case for change to PSR is built on fewer assets needed, not more. The biggest challenge with PSR, as I understand it, is the amount of process and policy change involved, including the impact on customer performance requirements.

By the way, I agree with Dealraker's views on BRK and its ownership of BNSF vs. the other Class 1's. I was just trying to figure out where/why the costs were increasing (and a presumably material intermodal customer lost) . Also, based on Harrison's record shown in the link from tedthedog, competing against properly executed PSR would seem to require a higher level of performance under the traditional system that BNSF continues to use. None of that is meant to be a comment on balance sheet management or other aspects of BNSF ownership.