Fannie and Freddie's apartment loans grow at nearly triple the pace of bank property lending

Put Trepp's 8.2% against the GSEs' $1.17 trillion and the year's net addition comes to roughly $89 billion ^[16]. The Federal Reserve's Z1 accounts show banks adding $88 billion over the same year, to a book of $3.1 trillion ^[9][1]. The figures come from three datasets (MBA for the balance, Trepp for the growth rate, the Fed for banks), and they set a multifamily rate against an all-property one, so the match is approximate ^[20][10][9]. Even so, two government-sponsored lenders grew their holdings by about as many dollars as the whole banking system ^[16][9]. Wolf Richter calls the government "the most aggressive CRE lender" ^[11].

That money goes into apartments. Multifamily is 46% of the $5.1 trillion income-producing stock and the GSEs hold half of it. Take 46% of $5.1 trillion, halve it, and you get the same $1.17 trillion, so by these figures the GSE book is apartment debt ^[2][19]. Add the $220 billion held by other government lenders and the public share comes to $1.39 trillion, or 27.2% ^[7][15]. Put the $771 billion in CMBS, CLOs and other securitizations on top, and the state and bondholders together hold 42.3% of income-producing CRE debt, against 37.5% for banks ^[6][18][4].

Whether this moves the low-rate era's losses onto taxpayers depends on what the new GSE loans are paying off. Richter wrote that "a property that made economic sense with a 3% loan didn't make economic sense with a 7% loan" ^[13]. The 8.2% could have been built three ways. One is refinancing apartment buildings at today's lower appraisals. Another is taking out finished projects from the $696 billion construction book, since construction loans are paid off when the completed property is refinanced ^[21]. The third is rolling pandemic-era loans at their old valuations. Only the third shifts an unresolved loss to the government, and a balance total looks the same in all three.

The bank record shows losses being taken. Bank CRE books stayed flat through 2024 and the first half of 2025 because banks sold troubled loans about as fast as they wrote new ones ^[8]. Richter wrote that the troubled loans ended in "sales of loans or properties at massive haircuts to new investors that then could do something with those properties because their cost basis was a lot lower and realistic" ^[14]. A seller that takes a haircut has booked its loss. Some of the office damage sits in the bank line. Foreign banks' US branches are counted there, among them Aozora, which disclosed big losses on a $1.9 billion US office-loan portfolio in February 2024 ^[3].

I think the evidence supports a narrower claim than a transfer: the government is lending into apartments faster than banks are lending into commercial property overall, while banks have absorbed losses on their worst loans ^{[10][9][8][14]}. The counter-case is that pandemic-era floating-rate loans became "economically impossible" when short-term rates began to soar in mid-2022. An 8.2% growth rate is exactly where a stale-value refinancing would be hard to see ^[12][10]. Richter writes that CMBS balances are surging again, though the figures in his article cover only the $771 billion held across CMBS, CLOs and other securitizations ^[11][6].