Bank vs non-bank: FX’s two-tier reality

Much has been said about the non-bank high-frequency trading (HFT) firms that supposedly lurk on the anonymous spot foreign exchange primary venues of EBS Market and Matching, waiting to pick off the unsuspecting banks managing client risk.

Some studies argue these firms can pull liquidity or widen spreads sharply in a crisis, a concern the Swiss National Bank raised this year.

But others argue non-banks contribute significantly to liquidity by passively trading against the pricing errors created by dealers’ aggressive flows, and supply the dominant share of price discovery through their high-frequency quote updates.

In an attempt to inform the debate, Paul Houston of CME Group and Simon Jones of LSEG gave a presentation to the European Central Bank’s foreign exchange contact group (FXCG) meeting in June, revealing fresh details on the participant mix on their venues, and on how non-banks and bank market-makers actually trade.

The presentation first looked at who makes liquidity. On LSEG Matching, roughly 67% of liquidity providers (LPs) were banks and 33% non-banks as of June 2026, up sharply from January 2022 when non-banks accounted for just 20%.

On EBS, the split is about 60% bank and 40% non-bank – a mix shaped by the access rule for its EBS Live Ultra data feed, which requires participants to make markets for at least 40% of their weekly volumes and contribute a minimum of $200 million in daily flow.

That maker/taker balance moves sharply when volatility spikes. On LSEG Matching, banks run an average ratio of 55:45; on March 3, in the opening days of the Iran conflict, it jumped to 70:30. However non-banks barely flinched, with their ratio only moving from 33:67 to 30:70.

The number of non-bank market-makers on the primary venues has also thinned. EBS hosted roughly 90 in 2016, and that figure has since more than halved.

The attrition is concentrated among the smallest players. The population of non-banks trading up to $250 million has fallen by around 70% since 2016, while the share transacting more than $2 billion has held steady – evidence that the business is consolidating around a handful of large, well-capitalised firms.

The FXCG minutes echoed this, noting that non-banks have “become important players in FX spot markets, particularly in liquid G10 pairs, but their activity is increasingly concentrated among a small number of large firms”.

Citadel Securities, Jane Street, Optiver and XTX Markets are among the established market-makers, alongside HFT names such as Jump Trading, Virtu and Tower Research. Other firms, however, such as Flow Traders, are understood to have pulled back a little from FX market-making over the years.

The presentation also highlighted differences in quoting behaviour between the banks and non-banks. Median order life for banks was 290 seconds on EBS and 181 seconds on Matching, while for non-banks it was 2.95 seconds and 1.64 seconds.

A case study of the Bank of Japan’s intervention on April 30 showed that as US dollar/yen slid from 159.35 to 155.50, 385 pips traded on EBS and 98.4% of all available price points changed hands – a burst of activity that saw decade-high volumes of $77 billion of USD/JPY on the day.

Non-banks tracked spot tick by tick as the pair fell over roughly two hours. They continuously refreshed their quotes around the touch and cancelled within seconds, so their liquidity clustered at the top of book around the price.

Banks behaved differently. They provided sticky, longer-lasting liquidity spread across the order book that did not chase the move – and, tellingly, they were resting bids all the way down to 155.50 well before spot got there.

The minutes drew the obvious conclusion: a complementary two-tier structure has taken hold in spot, in which non-banks dominate rapid price formation while banks provide depth, absorb risk and intermediate client flows.

However, the minutes also noted that non-bank liquidity is “conditional”, typically confined to small trade sizes, and the FXCG flagged concerns about growing concentration, limited transparency for end-users, and how resilient the market would prove if one of the handful of dominant firms pulled back in a period of severe stress.

What the presentation did not show is the other side of that structure – the steady retreat of banks from the primary venues themselves.

For years, dealers have invested in internal matching engines that skew pricing to offset residual risk, reducing their reliance on the external market. The rise of pod shops and the systematic trading community has accelerated the shift, supplying a continuous stream of axed flow that lets dealers rotate inventory without ever touching a public venue.

In addition, market structure specialists have argued that the pool of genuinely tradable liquidity on public FX venues has thinned, as highlighted by the CME outage in November last year.

Nevertheless, the presentation may be useful for future research into the true mix of non-banks and bank liquidity in the FX market.

Editing by Lukas Becker