The Fed has lost money three years running — ~$114B (2023), ~$78B (2024), ~$20B (2025), ~$212B total (Wolf Street chart of Fed data, via Barchart 09/2026) — after $50–117B a year of profits from 2009 to 2022. The mechanism is a duration mismatch on the central bank’s own book: QE loaded it with long, low-coupon Treasuries and MBS, and it pays IORB and reverse-repo rates on the reserves it created to buy them. When policy rates went above the portfolio’s ~2% average coupon, the carry turned negative. The shrinking 2025 loss is the same thing in reverse: cuts plus low-coupon bonds rolling off.
Why it matters and why it doesn’t. It doesn’t threaten the Fed — losses accumulate as a “deferred asset” and are netted against future profits; a central bank can’t go illiquid in its own currency. It does hit the Treasury: remittances (the ~$100B/year the Fed used to hand over) stop until the deferred asset is repaid, which is a real fiscal cost with no line in the budget — one more quiet reason the Fiscal dominance incentive to keep rates low is stronger than the official narrative. Barchart’s 🚨🚨 frames it as a scandal; the chart is the price tag of QE, arriving late. Sibling of Paper losses are real when you need cash — the same unrealized-loss stock, on the one balance sheet where no trigger can force the sale.