Inside US Leveraged & Inverse ETFs

What these products actually hold, read straight from constituent-level basket data - and where issuers genuinely differ in how they build them, versus where the data simply can't see far enough to tell.

What's inside, in general

Every US leveraged/inverse ETF is built from the same three raw ingredients - a direct holding of a stock or an existing 1x ETF, a total-return swap with a bank counterparty, or an exchange-traded future - combined in different proportions depending on how much leverage the product needs and how the issuer prefers to source it. By count, most of these funds are single-name or single-index products; by dollars, a handful of mega-cap broad-index funds dominate.

Fund count and AUM by what the fund's composition actually breaks down to

Fold the unweighted single-stock funds in by name rather than by measured weight, and single-stock products are actually the largest category by count (219 of 305) - it's only broad-index funds that dominate by AUM, because a handful of them (UltraPro QQQ, Direxion Semiconductor Bull 3X) are each individually larger than the entire single-stock cohort combined.

How the exposure actually gets built

Splitting funds by which of the three primitives shows up in their disclosed composition gives a rougher, more honest picture than the taxonomy above - because for many funds, the only leg visible is the collateral sitting behind an undisclosed swap.

305 funds by combination of legs visible in composition data

The direct-only row is the single biggest fund count, but it's a mix of two very different realities: broad-index funds that hold most of their exposure honestly, and single-stock funds (see §04) that show a ~100% stock stub while the swap providing the actual leverage sits entirely off-page.

Four funds, four structures

Composition data, read at the leg level, is where the taxonomy becomes concrete. These four are chosen to span the space: a multi-counterparty swap ladder, an extreme small-fund case, an issuer whose leverage mechanic is simply invisible, and a fund that discloses all three primitives at once.

Comparing issuers

Before comparing how issuers build similar products, it matters enormously whether we can see it at all - and visibility splits almost exactly along issuer lines, not product lines.

Share of each issuer's funds with no usable exposure-leg weight, by fund count

*Subject to availability from issuer - summary reflects current Ultumus composition feed

The three legacy sponsors - ProShares, Direxion, BMO - are fully visible in this feed. The newer single-stock cohort, mostly filed under smaller trust administrators, is 70–100% dark at the swap-leg level. This is very likely a reporting-pipeline gap between administrators rather than a real construction difference - but it means most "same stock, five issuers" comparisons (NVDA, TSLA, and similar names are each offered by four or five sponsors) simply can't be run on weight data. Only two clean head-to-head pairs survive with real numbers on both sides:

Net read: where we can actually see both sides, issuers building the same single-stock exposure look nearly identical (a bare stock holding, leverage mechanic undisclosed either way). The one differentiator the data proves outright isn't “swap style" at all - it's ETF vs. ETN: a fund that owns real securities versus a note that merely references a basket to calculate a payment.

Product list

Every fund behind the tables above, one row each, sorted by issuer then AUM. The same data underlies the exported spreadsheet; search matches issuer, fund name, ticker, or primary underlying. Contact us for the full list at info@ultumus.com

Reading notes & known limits

1. Coverage is bimodal by administrator, not a timing lag

Of 791 US shareclasses meeting the leverage/inverse definition, only 305 report usable exposure-leg weight - and that split holds day over day (301 of 788 on 2026-08-11, 305 of 791 on 2026-08-12), so checking an earlier date doesn't recover more. The missing ~490 are concentrated in specific administrators (Tidal/Themes/T-REX/VolatilityShares/KraneShares), reporting weight only on cash collateral, never on the swap itself. Treat any statement about "how issuer X builds its products" as bounded by what that issuer's administrator discloses to this feed, not as a claim about the actual legal structure.

2. Swap weight is gross, not concentration

Individual swap legs can show >1,000% notional on small funds, netting against negative collateral legs back to ~100%. A single leg's weight is only meaningful as "which counterparty/instrument dominates," never as a directly comparable concentration percentage across funds.

3. "Single-stock-like" means structurally concentrated

The category counts distinct reference instruments (≤2), not literal company count. A broad-index swap fund whose swap leg and futures leg are named differently for the same index can land in this bucket too (e.g. SQQQ) - a definitional caveat rather than something to fix further.

4. Single-day snapshot, cross-checked

The 305-fund detail (taxonomy, replication method, examples) references the 2026-08-12 composition run. Confirmed against 2026-08-11 that the fund set and coverage ratio are stable day to day - not re-verified for every table on this page.

Source: Ultumus constituent-level composition and fund reference data · US-domiciled shareclasses flagged as short or carrying leverage other than 1x · 791 total, 305 with usable composition weight · snapshot 2026-08-12, cross-checked against 2026-08-11