Posted by Colin Lambert. Last updated: March 2, 2026
It’s always, well, sometimes, nice when a central bank agrees with you. That was the case going through the Swiss National Bank paper on the CME outage last year and its impact on FX market functioning – the findings were pretty much identical with thoughts provided here at the time, even though some of the headlines they generated were wide of the mark.
That said, and far be it for me to disagree with academic analysis (and we all know what’s coming now…) when it comes to talking about concentration risks, I am not sure the paper is looking in the right direction.
The broad findings of the paper correspond with my thoughts in the December column, the market functioned without issue, and the main victims were CME’s brokerage stream and some poorly-equipped “liquidity providers” (read recyclers). What I found ironic, however, was that the paper focused on trading-related issues, which didn’t really exist beyond a few non-bank players, and raised, but didn’t really address, the bigger issue of genuine vulnerability in the system in the post-trade.
The paper did state, “The outage illustrates the technological concentration risk that results from relying on shared infrastructure”, before going on to highlight prime brokers, but it didn’t really go much further, instead returning to market conditions (which we will get to shortly).
I would argue that the concentration risk is not at the prime broker level – most clients now have two – rather, as I have argued before, it is at the next level, where the PBs meet, mainly on Harmony. Equally, a concentration risk that isn’t mentioned (and probably because the paper is dealing with a platform outage) is settlement risk. The FX industry has been lucky that CLS has not had a similar issue (although we should also paraphrase golfer Gary Player, by observing the harder CLS works on its robustness the “luckier” it gets), should it happen though, the chaos would be severe.
Something that did grab my attention on the market functioning cited in the paper was the line “trading was clearly impaired” during the outage. It should be noted that it also states further on, the event “did not lead to a major breakdown of trading activity”.
Impaired conditions may have been the experience for a few, although I have struggled to find them. If you look at the client segment most likely to have concentration risk – the corporate world – it neither interacts that heavily with FX futures, or EBS Market. In the hedge fund and asset manager world, these firms are largely connected to multiple venues and tend to trade where the best price is in their size. If one platform goes down, they move on to another.
The derivatives industry showed us a couple of years ago how bad things can get when one concentrated point of risk fails, FX cannot afford the same issue
The paper also states that transaction costs rose as a result of the outage, there is some colour provided in the report, but for who? Pairs on Matching, with no equivalent CME FX futures contract, saw no change in spreads, but that those with corresponding futures saw spreads widen by more than double. The paper also notes that spreads widened on pairs where the primary and futures venues were unavailable (in other words CME and EBS pairs).
So far, so good in terms of the impairment argument, but is that really the case? Yes, spreads widened – they often do when uncertainty over execution rises – but that is a blunt measuring stick because it doesn’t take into account size or the people using the venues. One segment was “impaired”, the LPs, but were they? Banks, as the paper notes, leant into their internalisation programmes and traded normally, but the non-banks? Here the paper observes that the worst impacted LPs were non-bank firms, who largely operate where? CME FX futures and the primaries, and in small size.
If we are looking at a participant trading in one million units (or less), then yes, spreads probably did widen, but how many “clients” are actually using these venues as their primary (no pun intended) venue? These venues have come to be dominated by non-bank firms – my information at the time was that the wider spreads were basically felt the most by non-banks.
Beyond wondering whether we should care if professional trading firms struggle because spreads widened by a fraction of a tick and upset the arb programme, there are several questions from the paper that are worth discussing.
Firstly, what were spreads like down the ladder in larger size? What were they for five million for example in one clip? Given that pricing often comes from banks, I suspect, as the paper notes, these spreads were not really impacted at all. The paper reinforces the realities of the modern FX market where price formation is largely off venues dominated by non-bank players pricing in smaller amounts. This has led to more algos being used as trades that would once have been a simple price request, were broken down into ones and less to take advantage of the tighter spreads available on the primaries.
Is this a true reflection of liquidity, however? I would argue it is not, perhaps a better example would be, what will it cost to get five done in one clip? At least that way we take out some of the white noise provided by last look-wielding, information-gathering, firms with no interest in actually adding to market quality.
Secondly, should we even be referring to CME FX futures, EBS Market and LSEG Matching as “primary” venues? Much has been made about the drop in volumes at these venues over a long period of time but the last decade can also be informational.
EBS from 2016 to 2025 was down more than 50%, and CME, which is widely seen as a success story, only saw FX volumes in notional terms grow something like 1-2% (in contract terms it is up 14.2% 2016-2025, but new, smaller, contract sizes have been introduced in that period). LSEG spot FX volumes were also up, by 7.2%, over that decade.
The problem is, apart from CME, we don’t know where the volumes are being traded, certainly we know they are not all on the CLOB, and in the case of EBS, it’s not all spot because NDFs are included. This means that CME is the only real indicator of volumes over these models.
Informed sources tell me that CLOB volumes on EBS are down over the last 10 years, and similar information from LSEG says they are slightly down. Either way, in terms of where people trade, the influence of these venues has waned, but of course, this is not really about trading volumes, it’s about data – and here, they most definitely have an influence on the market, especially with the non-bank firms and smaller banks who are minimal internalisers.
So, given the steady-ish nature of activity, and the importance of market data, yes, I think we should continue referring to the three as “primary” markets, at least until everyone gets their market data off the blockchain at minimal cost!
A third question the paper puts into my mind – where is price discovery really happening? In terms of the non-banks, the paper makes it pretty clear it’s the primary venues, but the banks? Here I am less sure thanks to the high internalisation rates. Internalisation muddies the water a little, but the reality is the single dealers are de facto ECNs. If I were a client, I would aggregate a few streams and take it from there. Not only would I get a good picture of where the market is, I would know in that five million units I want to trade. The banks are still in competition, reject rates are pretty low to non-existent on many of these streams, and the pricing is keen, provided there is no pattern of sniping.
This can already be done of course, by adopting a multi-dealer platform. My understanding is that for most customers – and this includes a number of “taker” as well as “maker” banks, the spreads on these RFS-type venues are as tight, if not tighter, than on the primaries, why not just go there? One issue is the use of last look on some ECNs is pretty egregious, the other is the cost to price for the LPs – at some stage that will register in the pricing.
The fifth question from the paper, is there consolidation risk? Absolutely, but it is not in the trading piece of the workflow, it is pretty much all in the post-trade – and that has little or nothing to do with the platforms, the LPs or the customers, it is a market infrastructure issue that needs addressing.
The derivatives industry showed us a couple of years ago how bad things can get when one concentrated point of risk fails, FX cannot afford the same issue. That means finding the means and support for an alternative to CLS and Harmony. Solutions exist, but for them to be commercially viable, they have to garner more support. In reality, this is unlikely to damage the incumbents much at all – volumes continue to grow – but it could provide the industry as a whole with a much-needed fallback.
The alternative is the incumbents develop a stand-alone, isolated, complete, solution in case the primary technology malfunctions. This is again a cost, and I would argue I prefer these firms to continue to invest in their robustness and embrace the idea of a degree of competition. Were they to develop redundancy, alongside, robustness, that would also be good for the industry as a whole.
The trading side of the business functions perfectly well, even if one or two primaries disappear, can we say the same for some other functions?
Ultimately, therefore, the SNB paper raises some interesting issues, but doesn’t really address them. It does reinforce what I have often said, when the smelly stuff hits the fan, you know who the truly robust LPs are – and who the recyclers are! What should be clear from the CME outage, as was the case with previous issues at the other primary venues, is that one or in this case two, of them disappearing does not really upset market functioning, depending upon who you care about. I have read headlines about “FX futures dependency”, but this simply doesn’t exist outside of a few non-bank “LPs” – if it did exist, then the market would not have functioned.
If you worry about professional trading firms (and LPs more generally) then yes, the events of last November were a concern, but personally I don’t. These firms make good money, and fair play to them, but with one or maybe two exceptions, the non-banks are not adding to market quality in my view. If, like me, you see the foreign exchange industry serving as a vital function to keep the global economy moving, then you worry about corporates, asset managers, and to a lesser degree, hedge funds. Here, as we noted in December, and is reinforced by inference at least in the SNB paper, nothing really changed – and that is a good thing.
There are technology consolidation risks in FX, that is clear, but are we muddying the waters by including a discussion about trading venues in the debate? I think we are, because the trading side of the business functions perfectly well, even if one or two primaries disappear. Can we say the same for some other functions?
