Unrealized bond losses cost nothing until a trigger forces sales — then they are losses like any other. The trigger is what varies: collateral calls (UK LDI pensions, 2022 — BoE had to buy gilts), deposit runs (SVB and the US regionals, 2023 — FDIC), and for insurers, policy surrenders and loans when rates rise. Spiking yields sat at the heart of both precedents, both within five years.
The 2026 specimen: Japan’s four biggest life insurers carry ¥15T (~$96B) of unrealized losses on their JGB portfolios — the Bloomberg chart shows the stack climbing nearly monotonically from ~¥1T in 03/2024 to ¥15T in 06/2026 (Nippon Life the largest slice). Blokland’s headline conclusion — Japan’s interest expense jumps “from 2% to 10% of GDP if yields stay here” — is full-stock-repricing math, the Stock vs flow repricing error: with JGB average maturity around 9 years (memory figure), that cost phases in over a decade, and only if yields stay. The insurer chart needs no such inflation; the losses are on the books now, waiting for a trigger.
The US banks’ own print, read correctly: “$325 Billion in unrealized losses 🤯” (Barchart, https://x.com/Barchart/status/2087408577417753012) — but the FDIC chart under the caption shows the peak was ~$690B in 2022–23, so the scary number is half the peak and shrinking as low-coupon bonds age toward par (Check the stat against its own chart). That’s the other half of this note’s mechanism: without a trigger, paper losses quietly pull to zero at maturity. The stock of losses is a vulnerability, not a countdown.
Same fact, thinner costume: GoldTelegraph’s “BREAKING” version (https://x.com/GoldTelegraph_/status/2086142633097707849). Blokland sells the “this time is NOT different” doom frame; the chart survives without it.