How the FX Market Is Structured: Dealers, Interdealer Venues, ECNs and the BIS Triennial Survey

The foreign exchange market is not an exchange. It is a decentralised over-the-counter (OTC) network in which a few hundred dealer banks quote prices to each other and to their clients, trade among themselves on a small number of interdealer venues, and match most client orders internally against other clients before any of that flow reaches a public screen.

In short: according to the final results of the Bank for International Settlements (BIS) Triennial Central Bank Survey, OTC FX turnover averaged $9.51 trillion a day in April 2025, up 27% from 2022. About 47% of that was dealers trading with other dealers, 49% was dealers trading with “other financial institutions” (smaller banks, asset managers, hedge funds and principal trading firms), and under 5% was dealers trading with non-financial companies. Electronic methods account for roughly 58% to 59% of turnover, but the anonymous central order books that people imagine when they think of “the market” are only a slice of it: the BIS reports that large dealers matched more than 80% of customer spot trades on their own books in April 2025. Retail-driven trading, the part a forex signal reader touches, was about $242 billion a day, or 2.5% of the total.

This is the map that every other piece in our institutional section builds on. Last look, the fixing scandal and how retail brokers hedge all make sense only once you know who sits where.

The misconception: “forex is one big market where retail traders trade with the banks”

Signal sellers often describe forex as a single global pool in which a retail account in Toronto trades “alongside” central banks and hedge funds. That picture is wrong in a way that matters for execution and for honesty about costs.

There is no central exchange and no consolidated tape for spot FX. Prices are formed across many venues and bilateral relationships at once.
Access is tiered by credit. Who you can trade with depends on who will extend you credit, which is why prime brokerage exists.
A retail trader almost never meets a bank directly. A retail broker is usually the counterparty, and the broker’s prices are derived from quotes it receives from wholesale liquidity providers.

The BIS survey is the best public evidence for how this structure looks, so we start there.

The source: what the Triennial Survey measures

The BIS has run the Triennial Survey every three years since 1986 (for FX). The 2025 round took place in April 2025, with central banks and other authorities in 52 jurisdictions collecting data from more than 1,100 banks and other dealers. Three features of the method matter when you read the numbers:

It counts dealer turnover. Reporting dealers report what their sales desks traded, wherever the trade was executed. A trade between two non-dealers that never touches a reporting dealer is largely invisible.
“Net-net” removes double counting. When two reporting dealers trade with each other, both report the trade. The BIS corrects for this locally and across borders to reach the “net-net” headline. Country rankings use a “net-gross” basis, which removes only local double counting.
Revisions happen. The preliminary release on September 30, 2025 gave $9.6 trillion; the final figures, published with the December 2025 BIS Quarterly Review (Huang, Krohn and Sushko, “Global FX markets when hedging takes centre stage”), give $9.51 trillion. We use Table 1 of the BIS’s final detailed tables throughout.

The BIS notes that April 2025 was unusual: US tariff announcements early in the month triggered a dollar sell-off and a rush to hedge dollar assets. The Quarterly Review estimates the tariff news added roughly $1.5 trillion a day to the month’s turnover, including an estimated $1.1 trillion spike in spot. Treat 2025 as a high-volatility snapshot rather than a normal month.

What is traded

From BIS Table 1 (daily averages, April 2025, net-net basis):

Instrument	Daily turnover	Share of total	Change since 2022
FX swaps	$4,015bn	42.2%	+6%
Spot	$2,952bn	31.0%	+42%
Outright forwards	$1,747bn	18.4%	+51%
Options	$632bn	6.6%	+108%
Currency swaps	$164bn	1.7%	+33%
Total	$9,510bn	100%	+27%

The largest instrument is not spot. FX swaps, which pair a spot trade with an offsetting forward, are mostly short-dated funding and hedging tools for banks and institutional investors. The BIS notes that interbank swap trading stagnated after 2022 while client hedging with forwards and options grew. For anyone thinking about signals, the relevant market is spot, a little under a third of the total.

Currencies. Because every trade involves two currencies, shares sum to 200%. The US dollar was on one side of 89.1% of all trades in the final tables (the preliminary release said 89.2%), the euro 28.5%, the yen 16.9% and sterling 10.2%.

Who trades with whom

Counterparty of the reporting dealer	Daily turnover	Share
Other reporting dealers (interdealer)	$4,443bn	46.7%
Other financial institutions	$4,627bn	48.6%
of which non-reporting banks	$2,195bn	23.1%
of which institutional investors	$1,261bn	13.3%
of which hedge funds and principal trading firms	$760bn	8.0%
of which official sector	$138bn	1.4%
of which other (incl. retail aggregators)	$229bn	2.4%
Non-financial customers	$441bn	4.6%

Three points stand out.

Corporates are a small share. Non-financial customers, the companies paying for imports or repatriating profits, were 4.6% of turnover, down from about 5.7% in 2022 (our arithmetic from the BIS change figures). The BIS notes that some corporate flow reaches dealers through corporate financing arms, which are counted as financial institutions.
Smaller banks are the largest client group. Non-reporting banks (regional banks that buy FX from big dealers and resell it to their own customers) were nearly a quarter of all turnover.
“Other” includes the retail industry. The BIS reporting guidelines tell dealers to classify trades with “retail aggregators”, which they describe as platforms such as Oanda, FXCM, Saxo, Gaitame.com or Gain Capital, as trades with “other financial institutions”. The reporting dealer sees the broker, not the retail client behind it.

Worked example: why the country table adds up to more than the world

The country table shows where sales desks sit. Summing every country’s net-gross figure gives $12,557bn, which is $3,047bn (32%) more than the $9,510bn global total. The gap is not an error. It equals, almost to the million, the $3,046.8bn of cross-border interdealer trading in Table 1: when a London desk trades with a New York desk, both countries report it, and only the global net-net figure strips out one side. On the net-gross basis, the shares are:

Location of sales desk	Share of net-gross turnover
United Kingdom	37.8%
United States	18.6%
Singapore	11.8%
Hong Kong SAR	7.0%
Top four combined	75.2%

The lesson for reading any FX statistic: always check whether it is net-net, net-gross or gross, because the same market can look a third larger depending on the convention.

The tiers: how a trade actually happens

Professionals think of FX liquidity in layers. The BIS December 2025 Quarterly Review (Box B, by Krohn, Schrimpf and Sushko) gives a clean taxonomy, and its footnotes name the venues in each layer.

Tier 1: the primary interdealer venues

Two anonymous central limit order books (CLOBs) sit at the core of spot trading in the major currencies: CME EBS Market (EBS, now owned by CME Group) and LSEG Matching (formerly Reuters, then Refinitiv Matching). The BIS calls them the “primary venues”. Orders there are firm: once matched, the trade stands. Each venue has its home currencies. The Financial Stability Board’s September 2014 report on FX benchmarks records that the WM/Reuters fix took its data from Thomson Reuters Matching for seventeen of its 21 “trade currencies” and from EBS for the rest, and WM only added Reuters Matching data for the euro, yen, Swiss franc and rouble in February 2015. That is why these two venues sit at the centre of the fixing scandal.

Dealers use the primary venues mainly to lay off inventory they cannot match internally. That is why the BIS uses their volumes as a gauge of stress: in April 2025 they saw only a modest pickup, which the BIS reads as “strained but not stressed”.

Tier 2: multi-dealer and secondary platforms

Multi-dealer platforms let a client request quotes from several banks at once (RFQ) or stream executable prices from several liquidity providers. The BIS lists 360T, EBS Direct, Currenex, Euronext FX (formerly Fastmatch), FXall, Cboe FX (Hotspot) and FXSpotStream among the secondary venues. Some are anonymous order books, some are disclosed relationship venues, and several let liquidity providers apply last look (the right to reject a trade request after seeing it), covered in our last look piece. Box B of the Quarterly Review counts more than 15 multi-dealer platforms available to customers.

Tier 3: single-dealer platforms and direct streams

Each large bank runs its own electronic platform. The BIS names Barclays BARX, Citi Velocity, Deutsche Bank Autobahn and UBS Neo as examples. Clients can also receive direct price streams from banks and from principal trading firms (PTFs), non-bank market makers; the BIS says customers can connect to more than a dozen of them.

Tier 4: voice

Voice trading has not disappeared. Direct voice dealing and voice brokers remain important for large or bespoke trades and for derivatives.

What the execution data show

BIS Table 1 breaks turnover down by execution method:

Execution method	Daily turnover	Share of total	Share of spot
Voice, direct	$2,393bn	25.2%	21.1%
Voice, indirect (voice brokers)	$1,265bn	13.3%	4.7%
Electronic, direct (single-bank systems, direct streams)	$3,092bn	32.5%	43.0%
Electronic, indirect (multi-dealer and anonymous venues)	$2,443bn	25.7%	28.3%
Undistributed	$317bn	3.3%	3.0%

Electronic methods total 58.2% of all turnover on these figures (the Quarterly Review’s chart reports 59%, rounding on a slightly different basis) and 71% of spot. Voice remains dominant in swaps and options. Within electronic trading, the Quarterly Review’s breakdown shows anonymous venues, including the primary CLOBs, at about 10% of global turnover. The screen that many retail traders imagine is “the market” is a tenth of it.

Internalisation: why most client flow never reaches a screen

The single most important structural fact about modern FX is internalisation. When a dealer’s client buys euros and, a few seconds later, another client sells euros, the dealer can offset the two on its own book and never touch an external venue. It earns the spread on both trades and leaks no information to the market.

The BIS defines the internalisation ratio for its survey as:

internalisation ratio = 1 minus (customer spot turnover hedged on external markets / total customer spot turnover)

where “hedged on external markets” includes trading on EBS or LSEG Matching, on mid-venues, aggressive trading on venues such as Currenex or Cboe FX, CME futures hedges, and taking liquidity from another bank. Only principal e-trading desks report it.

Box A of the December 2025 Quarterly Review reports that internalisation ratios exceeded 80% across all currencies in major trading hubs, and were higher still for G10 currencies, with Asian centres showing the strongest gains. The BIS also notes that dealers increasingly move risk between their own affiliates: intragroup spot trades rose 128% between 2022 and 2025, against 40% growth in all spot trading with reporting dealers.

Worked example: how much customer spot flow reaches the interdealer market

Customer spot turnover in April 2025 (dealers’ spot trades with other financial institutions plus non-financial customers) was $1,623bn + $144bn = $1,767bn a day, or 59.9% of spot. If an 80% internalisation ratio applied to all of it, only $353bn a day would need to be hedged externally. At 90%, the figure falls to $177bn. The reported ratios cover principal e-trading desks in the major centres, so this is an illustration of scale rather than a measured total. But it shows why the BIS describes much of FX as “non-visible” trading, and why public interdealer volumes understate how much risk is actually changing hands.

Economists call the older pattern, where dealers passed unwanted inventory to each other in a chain, “hot potato” trading (Richard Lyons, Journal of International Economics, 1997). High internalisation means fewer hot potatoes. The BIS argues this helped the market absorb the April 2025 shock without dysfunction, while warning that a market where most trading is private depends on dealers’ willingness to keep warehousing risk.

Prime brokerage: renting a bank’s credit

A hedge fund or principal trading firm that wants to trade with many banks would normally need a credit line with each. Prime brokerage solves this. The BIS guidelines describe prime brokers as large, highly rated banks that let clients trade, within credit limits, with a set of third-party dealers in the prime broker’s name. The client trade is “given up” to the prime broker, which becomes the counterparty to both sides.

The scale is striking. BIS Table 1 records $2,163bn a day of prime-brokered turnover (22.7% of all FX) and $1,281bn in spot alone, 43.4% of global spot turnover. Prime brokerage is how non-bank firms gained access to the interdealer venues, and how PTFs became major liquidity providers. Its credit mechanics, and what happened to it after the Swiss franc shock, are covered in FX prime brokerage explained.

Where retail sits

The BIS asks reporting dealers to flag retail-driven transactions: speculative, leveraged, cash-settled trades initiated by individuals. Retail flow reaches dealers in two ways, according to the reporting guidelines:

Directly, when an individual trades on a dealer’s own margin platform (counted under non-financial customers).
Indirectly, through a retail broker or aggregator, which then trades with the dealer as a wholesale counterparty (counted under other financial institutions).

In April 2025, retail-driven turnover was $242bn a day, 2.5% of the total, of which $109bn was spot (3.7% of global spot). Because retail brokers net client positions against each other before hedging, much retail trading never reaches a reporting dealer at all, so these figures describe only the part that brokers pass on. The mechanics of that netting are in our piece on how retail brokers hedge.

The practical consequence for anyone reading a signal: your “fill” is the broker’s price, which is derived from the broker’s liquidity providers, which in turn reflect their internal books and, ultimately, prices on EBS, LSEG Matching and the secondary venues. Every layer adds a spread and a small delay. There is no direct line from a retail platform to the interbank screen.

Who sets the rules

Spot FX is lightly regulated compared with exchange-traded markets. In most jurisdictions spot is not a “financial instrument” in the securities sense, so there is no FX equivalent of Regulation NMS or a public trade tape. Conduct is governed by a mix of instruments:

The FX Global Code. Launched in May 2017 after work by central banks and market participants under the BIS Markets Committee, and maintained by the Global Foreign Exchange Committee (GFXC). It has 55 principles across ethics, governance, execution, information sharing, risk management and settlement. The current version is dated December 2024; the GFXC’s review amended five principles (9, 10, 35, 50 and 51), mainly to improve transparency about client data and delegated execution and to add a “waterfall” of preferred settlement methods to reduce settlement risk.
Regulators’ recognition of the Code. The UK FCA recognises the Code under its scheme for unregulated markets. In its November 19, 2021 statement it said that behaviour in line with a recognised code tends to show that a person subject to the Senior Managers and Certification Regime is observing proper standards of market conduct.
Enforcement through other hooks. Where spot is not directly regulated, misconduct is reached through banking law, conduct rules for authorised firms, antitrust law, or (in the US) the link between spot benchmarks and regulated futures and swaps. The 2014 and 2015 fixing cases used all of these.

What can go wrong with this structure

Fragmentation. Liquidity is spread across dozens of venues and bilateral streams, so no participant sees the whole market. Professionals use aggregators to combine streams, at the cost of complexity.
Information leakage. Spraying a large order across many liquidity providers can reveal intent to all of them at once. This is one reason last look and internalisation exist.
Non-firm liquidity. Quotes subject to last look can be rejected, so displayed depth may overstate executable depth in fast markets.
Benchmark concentration. Huge volumes concentrate around fixing times, which creates incentives for manipulation, as the 2014 cases showed.
Credit chains. Prime brokerage concentrates counterparty risk in a few banks; when a client fails, the prime broker absorbs the loss.
Settlement risk. Two currencies are paid in different systems at different times. The 2024 Code amendments to Principle 35 reflect the GFXC’s view that “a material amount of unmitigated FX Settlement Risk” remains.

How professionals use the map

An institutional desk choosing where to execute does not ask “which broker has the best price?” in isolation. It asks a series of structural questions:

Which liquidity providers stream to us, and on what terms? Each has a disclosure cover sheet under the FX Global Code describing its capacity (principal or agent), last-look practice and use of client data.
Which venues do we need? Anonymous CLOBs for firm liquidity and price discovery, disclosed venues for relationship pricing, single-dealer platforms for large risk transfers.
How much information do we leak? Measured through post-trade markouts, the price move after each fill, by venue and counterparty.
How do we execute benchmark and large orders? Increasingly through algorithms rather than risk transfer, covered in FX execution algorithms.

None of this machinery is visible from a retail platform, and none of it is a trading edge. It is the plumbing that determines what a quoted price actually means.

The short version

The FX market is a decentralised OTC network, not an exchange. In April 2025 it turned over $9.51 trillion a day according to the final BIS Triennial Survey tables: 42% FX swaps, 31% spot, 18% forwards. Dealers trading with each other made up 47% of turnover, financial clients 49%, and corporates under 5%. Two anonymous order books, CME EBS Market and LSEG Matching, anchor interdealer spot trading, surrounded by multi-dealer platforms, single-dealer platforms, direct streams from banks and principal trading firms, and voice. Large dealers now match more than 80% of customer spot flow internally, prime brokers intermediate 43% of spot, and retail-driven trading is about 2.5% of the total. Conduct is governed mainly by the FX Global Code rather than market-structure regulation. This is a description of how the market is organised, not a recommendation about how or where to trade.