Thirteen Percent of the Trading, One Percent of the Assets
Commodity ETFs in the overnight tape: 74 funds, and three commodities that are 94 percent of it.
Sapinover ResearchAugust 20, 20268 min read
01
The Complex
Across the 60 sessions ending 2026-08-14, 74 commodity exchange-traded products printed on the overnight tape, for $7.16B in notional. That is 1.98% of a $362.14B tape spanning 3,148 symbols.
Two percent sounds like a rounding error until you look at how it is distributed. GLD, SLV, USO alone are 65.8% of it, and the single largest fund is 0.49% of the entire overnight tape by itself.
Before any of that means anything, one split has to be made and the fund categories do not make it. 42 of these funds hold the commodity, a futures position or a physically-backed trust, and account for $6.21B. The other 32 hold the shares of companies that produce it, for $943M. A gold miner carries equity beta, an operating cost base and single-name risk. It is not gold. Treating the two as one complex is the most common error in commodity-ETF commentary and it inflates the apparent size of the category.
02
Three Commodities
Sorted by overnight notional, with the share of each complex that is the commodity itself rather than producer equities.
Figure 1 · Where the money is
Overnight notional by commodity, each bar split between exposure to the commodity itself and exposure to the companies that produce it.
Bars are scaled to gold, the largest complex. Commodities below a quarter of a percent of the total are omitted from the chart and appear in the table below.
Commodity
Funds
O/N notional
Share
Is the commodity
Median frequency
Gold
22
$2.84B
39.7%
80%
42%
Crude oil
11
$1.97B
27.6%
87%
25%
Silver
8
$1.91B
26.7%
99%
60%
Energy sector
5
$140M
2.0%
0%
7%
Natural gas
4
$139M
1.9%
100%
98%
Copper
4
$70M
1.0%
5%
25%
Other
8
$52M
0.7%
98%
3%
Uranium
3
$20M
0.3%
0%
25%
Metals and mining
3
$9M
0.1%
0%
7%
Agriculture
4
$3M
0.0%
100%
8%
Broad basket
2
$1M
0.0%
100%
4%
Gold, Crude oil, Silver are 94% of overnight commodity notional. Everything else, agriculture and broad baskets and uranium and the rest, is rounding. There is no diversified commodity trade happening at night. There are three trades.
03
A Miner Is Not the Metal
The bullion-versus-producer split is not uniform, and the variation is the most useful thing in the table above.
Silver is the purest expression in the complex: 99% of silver overnight notional is the metal. Crude is 87% the commodity. Gold is 80%.
Copper inverts completely. Only 5% of copper overnight notional is copper. The rest is mining equities. Anyone reading copper in the overnight tape as a macro signal on the metal is, in practice, reading the share prices of a handful of miners with all the equity risk that carries.
The categories themselves will not tell you this. Three funds in our set sit in a direct commodity category while tracking gold miners: GDXU, GDXD, GDXY, together $160M. Two of them are exchange-traded notes rather than funds, which adds issuer credit risk on top of the equity risk. We reclassify by fund name rather than trusting the label.
04
Two Dozen Names
Frequency here means the share of available sessions on which a fund printed at all. The distribution is not a curve. It is two clusters with almost nothing between them.
Figure 2 · Two clusters, not a curve
Every fund placed by how often it printed against how much it traded. The vertical axis is logarithmic because notional spans four orders of magnitude.
The right edge is dense and the left edge is dense. The middle is nearly empty, which is what makes this a two-dozen-name market rather than a long tail of thin liquidity.
Prints on
Funds
O/N notional
Share of notional
every session
23
$6.79B
94.8%
most sessions
6
$59M
0.8%
intermittent
16
$303M
4.2%
episodic
29
$6M
0.1%
23 funds print on essentially every session and carry 94.8% of the notional. 29 funds print on fewer than one night in ten and carry 0.1%. Overnight commodity trading is a two-dozen-name market with a long tail that exists on paper and barely trades.
That shape matters for anyone sizing overnight liquidity. The tail is not thin liquidity. It is absent liquidity, most nights, in most of these products.
05
The Leverage Gap
Restricting to direct commodity exposure and removing the miner-tracking notes leaves 39 funds and $6.05B.
Figure 3 · The gap
Levered and inverse funds, measured two ways against the same set of direct-commodity funds.
Assets are an imperfect denominator for expected trading, since high-turnover products exceed their asset weight in any session. Section 07 takes that objection seriously.
Structure
Funds
O/N notional
Share
Median frequency
Median % of ADV
unlevered
30
$5.24B
86.5%
27%
0.19%
long levered
5
$388M
6.4%
98%
1.60%
inverse
4
$428M
7.1%
99%
1.36%
Levered and inverse funds together are 13.5% of direct-commodity overnight notional while holding 1.1% of the assets in that set. That is roughly a twelvefold over-representation.
The behavioural gap is wider than the size gap. Unlevered commodity funds print on a median 27% of sessions. Long levered funds print on 98% and inverse funds on 99%. On participation, the median unlevered fund trades 0.19% of its regular-session volume overnight while the median levered fund trades 1.60%, an order of magnitude apart.
The direction of the levered exposure is worth noting too. The inverse bucket is slightly larger than the long-levered bucket, $428M against $388M, across fewer funds. Whatever is happening overnight in levered commodity products, it is not a one-way book.
The ratio below is overnight share volume on one ATS against the fund's average regular-session daily volume across all venues. The numerator is a single venue at night and the denominator is the whole market by day, so it is a floor rather than a share of total trading.
Figure 4 · Structure sorts the field
Overnight participation against trading frequency. Colour is the fund's structure, not its commodity.
The levered and inverse funds occupy the upper right almost exclusively: they print nearly every session and take a far larger share of their own daily volume overnight than unlevered funds do.
The ceiling sits around three to four percent. Given the asymmetry in the ratio, the true overnight share across all four ATSs trading the session is higher than this, and for the most active names it is likely to be meaningfully higher.
07
Where We Might Be Wrong
The counts are reproducible from the published table. The readings of them are not, so here are the objections we take seriously.
One venue is not the session. This is BlueOcean only, one of four NMS Stock ATSs trading overnight. If commodity flow is distributed differently across those venues than equity flow is, the shape here is a sample rather than the market, and the concentration in particular could look different on a combined tape.
The leverage gap may be a size artifact. Levered products are small and turn over quickly by construction, which is what they are built to do. Comparing trading notional to assets under management will always flatter a high-turnover product, in the day session as much as at night. The finding is that the ratio is extreme overnight, not that it is unique to overnight, and we have not measured the same ratio in the day session to prove the gap is specific to the overnight window.
The denominators are ten weeks older than the numerators. Assets and average daily volume come from a June snapshot; the tape runs to August. Fund assets move with both flows and prices, and precious metals in particular moved over that window. The participation ratios and the assets comparison inherit that lag.
Frequency is not depth. A fund that prints every session might print one small trade each night. We report how often a fund appears and how much it traded, but a single print and steady two-sided liquidity look identical in this measure.
##
Method and Sources
The table. All 74 funds with classification, notional, volume, session count, frequency and participation, as a CSV.
How this was built. Fund classification comes from the ETF enrichment dataset. Four categories are treated as direct commodity exposure and three as equity proxies, then every fund is re-checked against its own name so that miner-tracking products sitting in commodity categories are moved. Overnight activity is the BlueOcean tape over the 60 sessions ending 2026-08-14. Frequency is sessions with a print divided by sessions available. Participation is average overnight share volume divided by the fund's average regular-session daily volume.
One correction worth recording. Inverse funds carry the leveraged flag as well as the inverse flag. An initial pass therefore counted them in both buckets and reported the entire levered total as long-levered exposure. The published split excludes inverse funds from the long-levered bucket, which is why long levered and inverse are reported as 5 and 4 funds rather than 9 and 4.
Vintage. Notional, volume and frequency are current to 2026-08-14. Assets, expense ratios and average daily volume come from the ETF enrichment snapshot generated June 8, 2026. Any figure combining the two, meaning the participation ratios and the assets comparison in section 05, carries that gap.
