Dark Pools Explained: What FINRA's Own Data Shows in 2026

Dark pools are private trading venues where buy and sell orders are matched without displaying quotes to the public market. They are legal, regulated as Alternative Trading Systems (ATS) by the SEC, and every one of them reports its volume to FINRA on a weekly basis. In the week beginning 10 August 2026, US dark pools printed 11.76 billion shares across 163.2 million trades, worth roughly $871.7 billion — about 2.35 billion shares a day.

The part almost nobody gets right: dark pools are the smaller half of off-exchange trading. In that same week, off-exchange volume that did not go through a dark pool came to 28.0 billion shares. Dark pools were 29.6% of all off-exchange activity. The rest is wholesaler internalisation — the retail order flow that never touches a lit exchange or a dark pool.

What a dark pool actually is

A lit exchange publishes its order book: you can see resting bids and offers before you trade. A dark pool does not. Orders sit unseen and match against each other, usually at or inside the National Best Bid and Offer. The trade is reported publicly after it happens, through a FINRA Trade Reporting Facility, so the tape is complete — the pre-trade quote is what is missing, not the print.

The reason institutions use them is market impact. If a pension fund needs to sell two million shares and shows that on a lit book, the price moves against it before the order is filled. Hiding the intention is the entire product.

Three kinds of venue get lumped together

• Broker-dealer pools run by large banks, matching client and sometimes principal flow (UBS ATS, Sigma X2, Level ATS).
• Independent / agency pools with no house position (Intelligent Cross, PureStream, BIDS).
• Wholesaler internalisation — technically off-exchange but not an ATS, and this is where most retail market orders are filled.

The numbers: week of 10 August 2026

All figures below come from FINRA's own OTC Transparency dataset for NMS-listed stocks (UTP and CTS tapes combined), published 14 September 2026.

Category	Shares	Trades	Average print
Dark pools (ATS)	11.76 bn	163.2 m	72 shares
Off-exchange, non-ATS	28.00 bn	129.3 m	216 shares
Total off-exchange	39.75 bn	292.5 m	—

Read the last column twice. The average dark-pool print was 72 shares. The folklore version of a dark pool — a quiet room where blocks of half a million shares change hands — describes a small minority of the activity. Most dark-pool volume in 2026 is algorithmic child orders, sliced small, hunting for price improvement a fraction of a cent inside the spread. Non-ATS off-exchange prints were three times larger on average.

Dark-pool share of off-exchange volume, 2026

Week starting	ATS (bn shares)	Non-ATS (bn)	ATS share
25 May 2026	11.08	30.62	26.6%
29 Jun 2026	11.53	29.58	28.1%
20 Jul 2026	11.71	29.71	28.3%
27 Jul 2026	14.42	31.09	31.7%
3 Aug 2026	13.91	32.97	29.7%
10 Aug 2026	11.76	28.00	29.6%

The range is tight: 26.6% to 31.7% across the summer. Dark-pool share creeps up in heavier weeks — the 27 July week was the busiest of the six and also the highest ATS share, which is what you would expect when institutions are the ones doing the trading. It is a stable structural feature, not a trend anyone should be extrapolating.

Which dark pools are actually busiest

FINRA reports ATS volume venue by venue, per symbol. Here is the week of 10 August 2026 for three of the most heavily traded names in the market. Thirty-plus venues printed each of them; these are the leaders.

	SPY (48.9 m ATS shares)	NVDA (71.4 m)	TSLA (21.3 m)
1	Dealerweb — 39.8%	UBS ATS — 13.7%	UBS ATS — 18.2%
2	UBS ATS — 10.4%	Intelligent Cross — 10.7%	IBKR ATS — 14.4%
3	Intelligent Cross — 9.7%	Level ATS — 7.3%	Intelligent Cross — 10.8%
4	Virtu MatchIt — 6.7%	Sigma X2 — 7.3%	Sigma X2 — 8.7%
5	Level ATS — 4.9%	IBKR ATS — 6.5%	Level ATS — 7.3%

Two things stand out. Dealerweb's 39.8% dominance in SPY does not repeat in the single stocks — that is ETF-specific institutional flow, not a general ranking. And no venue owns more than about 18% of a single stock: this is a fragmented market, which matters because it means no one dark pool sees enough of the picture to "know" where a large order is going.

What retail traders get wrong about dark pools

"Dark pool prints predict the next move"

A dark-pool print is a completed trade reported after the fact, with no side attached — you cannot tell from the tape whether the initiator was buying or selling. Services that sell "dark pool buy signals" are inferring side from price relative to the mid, which is a guess with a coin-flip failure rate on small prints. Given the 72-share average print, most of what shows up in a dark-pool feed is algorithmic noise.

"My order gets filled in a dark pool and that's why I lose"

Retail market orders in the US overwhelmingly go to wholesalers, not dark pools — that is the 28 billion non-ATS shares in the table above. The debate worth having about retail execution is about payment for order flow and price improvement, not about ATS venues.

"Dark pools hide volume from the market"

They hide the quote, not the trade. Every dark-pool execution is printed to the consolidated tape and folded into published volume. What is genuinely delayed is venue attribution: FINRA publishes ATS data with a two-to-four week lag for Tier 1 stocks, which is why the freshest data in this article is from mid-August.

Does any of this change how you trade?

For most people, honestly: no. Dark pools are plumbing. But three practical points fall out of the data.

• Liquidity in a name is deeper than the lit book suggests. A large share of US equity volume is printed off-exchange in a typical 2026 week; a thin-looking order book on a large-cap is not the whole market.
• Volume spikes need context. When a stock's reported volume jumps, part of that can be off-exchange reporting rather than a lit-market stampede. Use it as a reason to check the news, not as a signal by itself.
• Treat "dark pool data" products with scepticism. The underlying FINRA dataset is free, lagged, and side-less. Anything sold as a real-time dark-pool edge is repackaging the tape.

If you are building a view on index-level positioning, our S&P 500 forecast for 2026 covers the flows and valuation picture, and the volume distortion around quadruple witching dates is a far more reliable calendar effect than anything in the ATS feed. For ETF execution specifically, the spread and liquidity comparison in VOO vs SPY matters more to your fill than which venue prints the trade.

Bottom line

Dark pools are a real, large, well-documented part of US market structure: 11.8 billion shares and $871.7 billion in a single August 2026 week. They are also consistently under a third of off-exchange trading, dominated by small algorithmic prints rather than blocks, and fragmented across dozens of venues with no single dominant player. The data is public and free from FINRA. The mystique is mostly marketing.

Traders who want to follow index and equity setups in real time can see how we structure ours on the indices signals hub, or open a trading account to access live market depth. Nothing here is investment advice; trading carries risk of loss.