BMO and RBC sold Canada's largest payment processor for about 57 cents a transaction

Divide CA$2bn by the 3.5 billion transactions Moneris says it runs each year and you get about 57 Canadian cents of purchase price per annual transaction ^[9], which is useless as a multiple, since the report carries no revenue or earnings figure and no terms for the referral agreements ^[16], and quite useful as a description of what Francisco Partners believes it bought: an installed base that Abhijit Chakravarty of IBS Intelligence calls extraordinarily difficult for a fintech to assemble from scratch, whose upside is software, data and newer commerce capabilities bolted onto merchants who are already connected ^[14]. (American Banker prints the US-dollar equivalent of the price as $1.4 million ^[12], which cannot be right, so work in Canadian dollars.)

The capital line is the one worth slowing down for. Each bank takes CA$1bn ^[17], and BMO guided to roughly 15 basis points of CET1 improvement ^[6], but 15 basis points is a ratio, so unless BMO's risk-weighted assets happen to run near CA$667bn, the capital actually credited is a good deal less than the gross CA$1bn ^[18]. That gap is what carrying value already inside the ratio, tax on the gain and any unwinding deductions normally account for; it is ordinary, and it is also the reason CA$1bn of proceeds and 15 basis points of CET1 should never be quoted as the same benefit.

What the banks kept is a referral pipe with no published price. James Hicks, Moneris's chief executive, framed the long-term referral agreements and continuing commercial relationships as a foundation for continuity ^[3], and Chakravarty's reading is that ownership of the infrastructure has been separated from distribution of it, which raises the question of what technology a bank needs to own at all ^[13]. In cash terms the swap is legible enough: surrender consolidated processing economics and the standing obligation to fund a processor's roadmap, collect CA$1bn now plus an undisclosed referral stream, and spend the difference on software and AI, which is where Chakravarty says bank money wants to go ^[15]. A referral fee is a fraction of a merchant's processing spend, and a fraction beats the whole only when the fraction arrives without capex.

The tidy version of this story is that rails are rentable and relationships are not, and I half believe it, but the counter-thesis sits inside the same article twice. BMO's CFO Rahul Nalgirkar filed Moneris alongside two other disposals in 12 months of businesses that missed the bank's growth and return-on-equity bar ^[8], which is portfolio pruning rather than a doctrine about payments. And Capgemini's Michael Levens describes banks travelling the other direction, monetizing modernized infrastructure by commercializing it or spinning it into a separate entity ^[11], with Fifth Third's Newline as this year's example of a bank becoming the plumbing instead of selling it ^[10]. Both can be true at once; they are not the same template, and only one of them is available to a bank that never incubated a processor.

Being wrong here looks like one of two things. Francisco Partners layers software over 3.5 billion transactions, doubles the asset, and 57 cents a transaction reads as a cheap exit from something that was compounding. Or the referral economics prove thin enough that what the two banks actually bought with a 25-year-old business ^[2] was 15 basis points and a lighter technology budget, which is a real trade, or rather the more interesting version of the trade, but not evidence that distribution was the valuable half.