Every fund that enters a swap has to say, in Form N-PORT, who is on the other side of it: the counterparty's name and its legal entity identifier, line by line. Read one fund's filing and that is a list of banks. Read all of them and sort the funds by how much of the fund is actually swaps, and something turns up that is the opposite of what the shape of the risk would suggest.
The funds that exist because of swaps spread them. The funds that barely use swaps do not bother.
| Swap book, as a share of net assets | Funds | Banks used, mean | Largest bank's share, mean |
|---|---|---|---|
| 100% or more — the swaps are the fund | 43 | 6.4 | 42.2% |
| Under 5% — the swaps are a rounding error | 61 | 2.1 | 84.6% |
Read from each fund's own most recent Form N-PORT · 137 funds name a counterparty · counterparties resolved on the filed legal entity identifier, not on the spelling.
The leveraged funds, which have everything to lose
TQQQ is 256.3% of its net assets in swap notional. It is not a fund that uses swaps; it is a fund that is swaps, with a share class attached. Lose a counterparty and there is no fund left to reconstruct.
So it uses ten banks, and the largest of them carries 12.4%.
| Fund | Swap notional, % of net assets | Banks | Largest bank | Its share |
|---|---|---|---|---|
| TQQQ | 256.3% | 10 | Citibank NA | 12.4% |
| SOXL | 243.3% | 8 | Citibank NA | 26.2% |
| SPXL | 229.6% | 7 | Citibank NA | 19.2% |
| QLD | 122.3% | 10 | Goldman Sachs International | 27.6% |
| SSO | 120.4% | 10 | Goldman Sachs International | 37.6% |
This is what counterparty risk management looks like when the counterparty risk is the whole business. Nobody in that table is betting the fund on one phone number.
The plain index funds, which have almost nothing at stake
VOO holds a swap book too. It is 0.2% of the fund — cash management, index exposure on money that has not settled yet, the sort of thing that exists so the tracking error does not drift on a Tuesday.
Bank of America is 95.9% of it.
| Fund | Swap notional, % of net assets | Banks | Largest bank | Its share |
|---|---|---|---|---|
| VTI | 0.170% | 3 | Bank of America NA | 97.7% |
| VOO | 0.200% | 4 | Bank of America NA | 95.9% |
| VUG | 0.040% | 1 | Bank of America NA | 100.0% |
| VB | 0.030% | 2 | Bank of America NA | 95.4% |
| VTV | 0.480% | 3 | Bank of America NA | 74.4% |
Thirty-one of the 137 funds put 100% of their swap book with a single bank. Twenty-two of those thirty-one have a swap book smaller than 1% of net assets.
This is not a warning
It would be easy to write the second table as an alarm, and it would be wrong. A 100% concentration in an exposure worth 0.04% of the fund is not a risk; it is an administrative convenience. VUG's entire swap book is $147 million against $379 billion of net assets. If Bank of America vanished tomorrow, the arithmetic that would reach a VUG holder rounds to nothing.
The honest reading is the reverse of the scary one: concentration tracks how little is at stake, not how much. When a fund has something to lose it diversifies; when it has nothing to lose it signs with whoever it already has paperwork with. Both are rational. Neither is visible on a fund page, a fact sheet or a holdings screen, because none of them carry the counterparty at all.
What is worth knowing is the first table, not the second: that an instrument sold to retail as a day-trading product is, underneath, a more carefully spread book of bank exposure than the index fund sitting next to it in the same brokerage account.
Where this comes from, and what was wrong with it until today
These figures are computed from Item C.11 of each fund's Form N-PORT, where the counterparty name and LEI are filed per swap. The full dataset is published at /data/leverage.json under CC BY 4.0, with the method at /methodology/.
A correction belongs here. Until 2026-10-07 this dataset keyed counterparties on the name exactly as filed, and filers spell the same bank several ways — GOLDMAN SACHS INTERNATIONAL and Goldman Sachs International, BANK OF AMERICA, N.A. and Bank of America NA, BNP Paribas SA and BNP Paribas. Each spelling counted as a separate firm, so every bank with more than one spelling was understated, and most have more than one. What looked like 117 counterparties is 66. Bank of America appeared in 43 funds with $11.6bn; it is in 77 funds with $40.1bn.
The filings carry a counterpartyLei on every one of these lines, so the identifier is now the key and the name is only a label — the same rule this site already applied to securities-lending borrowers and should have applied here from the start. No fund's exposure figures moved. Only the question of who the exposure is to. The change is dated in the data dictionary and the tables above are computed after it.
Related: who borrows from index funds, the other side of the same question — the banks that borrow stock from funds rather than sell them swaps.