Bank of America's $90 Billion Bond Hole Is Now Testing the 'Buy the Dip' Pitch the Bank Is Selling Clients

Bank of America's bond portfolio has been a mark-to-market story for three years, but the Q3 update is the sharpest widening since the 2023 regional-bank crisis. The Q2 2026 10-Q filed July 31 disclosed $82,095 million in HTM gross unrealized losses plus $2,995 million on AFS — $85.1 billion total. Three-quarters of that pain sits in agency mortgage-backed securities: $379.2 billion of amortized-cost agency MBS now carries a fair value of $310.8 billion, a $68.3 billion hole on one security type. Add $13.0 billion of losses on $121.3 billion of Treasuries and the bank has $81 billion of the $85 billion already accounted for before the Q3 yield move.

The Q3 math is mechanical. The 10-year Treasury yield finished Q2 near 4.45% and traded above 5.00% intraday on September 16 per Federal Reserve H.15 data. A roughly 55-basis-point parallel shift on a portfolio with a duration north of six years adds the kind of mark-to-market pain Barron's estimates at "more than $10 billion" — pushing the total past $90 billion. The same yield move is what BofA's own strategists now call the market's biggest tail risk.

The single biggest line in Bank of America's loss stack is agency mortgage-backed securities, which is exactly where the post-2022 rate shock had to land. Of $505.8 billion of HTM debt securities at June 30, 2026, $379.2 billion (75%) is agency MBS, $121.3 billion is U.S. Treasuries and government agencies, and $5.4 billion is other taxable. The agency MBS block is $68.3 billion underwater — an 18% mark on a security type that pays roughly 2% to 3% coupons issued between 2020 and 2022. That long-duration, low-coupon structure is what makes the portfolio so rate-sensitive: every additional 25 basis points in the 10-year adds billions to the unrealized loss column.

The classification matters. HTM accounting lets BofA carry these bonds at amortized cost, so the $82.1 billion HTM loss never hits the income statement — it sits in a disclosure footnote. The $3.0 billion AFS loss, by contrast, flows through other comprehensive income and reduces reported tangible book value. Combined, the $85.1 billion loss is roughly 41% of BofA's $206.1 billion tangible common equity at June 30, 2026. That ratio is what makes the headline feel existential even though the bank's Q2 net income was $9.1 billion at a 17.0% return on tangible common equity.

Inside Bank of America, the bond trade is being marketed and warned against simultaneously. The September 15 BofA Global Research Fund Manager Survey shows a net 49% of respondents overweight global equities (down from 56% last month) and identifies a "disorderly rise in bond yields" as the market's largest tail risk. That survey was published the day after BofA Chief Investment Officer Chris Hyzy's Private Bank weekly note counseled clients to "continue to see any short-term market dips as potential opportunities to add strategically to your portfolio" and to treat higher rates as income opportunities for bond holders. Strategists are de-risking; the wealth desk is accumulating.

The split is not a contradiction in the wealth desk's defense — long-duration buyers at 5% yields are being paid more than long-duration buyers at 2%. The contradiction is that the same parent company holding $82.1 billion of HTM losses at June 30 cannot actually execute the trade it's recommending. Selling any meaningful slice of the HTM book would crystallize the loss; rolling the book into new 5%-coupon paper shrinks the balance sheet and the net interest income line that drove Q2's $15.997 billion. BofA is telling clients to do what it cannot do itself.

The FDIC's Q2 2026 Quarterly Banking Profile put industry-wide unrealized securities losses at $326.7 billion, barely changed from $325.1 billion the prior quarter. Bank of America's $85.1 billion share is 26% of the system total — far above its roughly 11% share of U.S. banking assets. The mismatch is the story: BofA is the industry leader in mortgage origination, mortgage servicing and securities warehousing, so its HTM book is structurally larger than its deposit or capital base implies. The peer set is exposed, but BofA is exposed hardest.

The 2025 peer disclosures are the most recent published snapshots and predate the Q3 yield move, so today's gaps are likely wider — but the rank order is the point. Wells Fargo is the second-most-exposed bank and shares the same agency-MBS concentration; JPMorgan Chase sits in the middle with the most diversified liability profile to offset mark-to-market noise; Citigroup has the smallest legacy mortgage book. Among regionals, U.S. Bancorp and other super-regionals carry similar duration exposure at a smaller absolute scale. The supply chain runs through one place: agency MBS issued by Fannie Mae and Freddie Mac, whose 30-year mortgage coupons from 2020–2022 are now trading roughly 200 basis points above their original yields.

Bank of America closed at $58.18 on September 18, down only 0.5% on the day and well off the September 14 post-Barclays low. The reason the stock hasn't capitulated is that Q2 2026 results, reported July 14, showed the franchise earning through the loss. Revenue of $31.6 billion was up 7% year-over-year; net interest income of $15.997 billion led the charge; net income of $9.1 billion translated to EPS of $1.21 (a 7.7% consensus beat) and a 17.0% return on tangible common equity. Investment banking fees jumped 50% to $2.1 billion, and Global Markets posted sales-and-trading revenue of $7.1 billion. The mark on the bond book is a footnote; the income statement is the headline.

The CEO's Barclays remarks on September 14 changed the setup, though. Brian Moynihan guided Q3 investment banking fees to $1.6 billion–$1.8 billion, down roughly 10% from the $2.0 billion the bank earned in the year-ago quarter and below the $2.1 billion Q2 just reported. The stock fell 5.1% that day to $59.47. The signal: after two strong quarters of capital-markets revenue, the deal pipeline is cooling at exactly the moment the rate environment is becoming more hostile to the balance sheet. Q3 2026 will be the first quarter where both the bond mark and a softening fee cycle show up at once.

The synthesis is uncomfortable. Bank of America's Q2 franchise is genuinely strong — 17% ROTCE, $16 billion of net interest income, double-digit IB growth — but the bond book is the largest single interest-rate bet in U.S. banking, sitting at $506 billion of amortized cost with $82 billion of paper losses and no realistic path to monetization without crystallizing. The bank is selling clients a trade it cannot execute at scale. That gap is the rate call, and it is the most important one for the rest of 2026.