Welcome to another issue of Net Interest, my newsletter on financial themes. To support my work and gain access to an extensive archive of 100+ past articles on financial markets, fintech and more, please join hundreds of others and sign up as a paid subscriber. This week, paid subscribers have also been sent comments on Liability Driven Investment, Mortgage Originators and Credit Suisse.
In 2016, an affiliate of high-speed trading firm, Jump Trading, made an unusual purchase: It bought a field in the city of Aurora, Illinois. There, on the side of a brick hut, it mounted a pair of small microwave antennas and pointed them eastwards. Jump’s offices were only 35 miles away, but the antennas reduced its communication time with the Chicago Mercantile Exchange (CME), whose data centre was situated across the street from the field in Aurora. The antennas would receive financial information from the data centre via a short cable and beam it to financial markets on the East Coast.
The field cost close to $14 million to buy but it allowed Jump Trading to shave a microsecond (one millionth of a second) off the time taken to exchange data with the CME. In the world of high-speed trading, that conferred a material advantage – even if it didn’t endure. Within a year, competitor DRW Holdings installed its own antenna on a light pole overlooking the data centre and McKay Brothers, a telecom company that sells bandwidth and fast data to the trading industry, installed a pole rigged with multiple antennas.
The episode is just one of many in the speed race that has underpinned financial markets for centuries. We’ve come a long way since the painfully slow process of hand-delivered notes. An eleventh century letter reveals the frustrations one Mediterranean trader suffered from such high latency:
The price in Ramle of the Cyprus silk, which I carry with me, is 2 dinars per little pound. Please inform me of its price and advise me whether I should sell it here or carry it with me to you in Misr (Fustat), in case it is fetching a good price there. By God, answer me quickly, I have no other business here in Ramle except awaiting answers to my letters… I need not stress the urgency of a reply concerning the price of silk…
Fortunately for later traders, entrepreneurs found ways to reduce the waiting time. As we discussed in a prior issue of Net Interest, the Rothschild family established a network of messengers throughout Europe to facilitate the flow of information: Nathan Rothschild was famously the first man in London to hear the news of Napoleon's defeat at Waterloo in 1815. Later, Paul Reuter, founder of the Reuters news agency, employed carrier pigeons between Brussels and Aachen to bridge a gap in telegraph stations on the route connecting Berlin to Paris.
Just as cables disrupted carrier pigeons, more recent innovations have been no less immune to disruption. Michael Lewis’s book Flash Boys opens with an account of a new high-speed fibre optic cable being laid between New York and Chicago. Unlike previous connections which followed railway tracks and wound their way round mountains, this one was dug in a nearly straight line. It cost $300 million to build and reduced the round trip communication time between the cities from 16 milliseconds to 13 milliseconds (for comparison, the blink of a human eye lasts between 100 and 400 milliseconds).
Yet within a few years, the new connection was obsolete. Shortly after it was finished, microwave technology took off with capacity to reduce transmission times further. Firms strung together chains of microwave towers to create a route that became 40% faster than the best fibre. Round trip times between Chicago and New York fell to 10 milliseconds and then 9 milliseconds; they are now sub 8 milliseconds. Capital expenditure of the original route was estimated at around $140 million, with an additional $20 million per year required to operate it.
For those high speed trading firms that could afford the expenditure, the returns were good. The CME trades over a quadrillion dollars notional per year. Suppose only a tenth of that is correlated enough with markets in New York or Europe to allow arbitrage trading opportunities, that’s still a big number. Assuming the relevant CME contracts have an annual volatility of 10%, then, according to one calculation, around $200 million of aggregate value per year can be captured in the 2.5 milliseconds that wireless networks save over fibre (one way).
For a while, high-speed trading outfits reaped these returns. But in the years following Jump Trading’s effort to squeeze an additional microsecond out of its process, two things happened. First, asset price volatility declined. Through to the end of 2017, volatility – as measured by the VIX Index – was on a downward trend. Even after it troughed, it wasn’t until the onset of the pandemic that it came out of the low range it had been in since around 2012. Low volatility reduced the revenues firms could extract out of high-speed trading. When Virtu Financial went public in 2015, it was off the back of adjusted net trading income of 33 cents per trade the year before; that number only went down.
Second, incremental returns from speed diminished as transmission times reached their Einsteinian limit. It’s simply not possible to transmit data between Chicago and New York any faster than 3.94 milliseconds one way. The competitive advantage of pigeons over men and fibre over copper is clear, but each subsequent iteration has lower impact. The closer you get to the speed of light, the less there is to squeeze out.
In response, the high speed trading industry consolidated. Firms such as Getco, Chopper, Infinium, Teza, RGM Advisors and Sun Trading disappeared through merger and the survivors scaled up.
Yet in the midst of this consolidation wave, a new firm was born – XTX Markets. Based in London, the firm was spun out of hedge fund GSA Capital in 2015. Today, it is one of the largest trading firms in the world, active across foreign exchange, equities, commodities and options, where it trades over $295 billion of volume per day. In 2021, it reported $2.0 billion of net revenue across three UK-registered legal entities, on which it earned $920 million of net profit. With volatility higher in 2022, it is likely to beat those numbers this year. 1
Like other trading firms, XTX makes money via repeated small gains on trades. It is both a price ‘maker’ and a price ‘taker’ and competes with others on both fronts. Price makers make markets by continuously posting bids and offers in order books for others to execute against; price takers execute against orders that are already present in order books. One way firms compete is by exploiting their speed advantage. Price takers can pick off stale quotes put up by price makers if they can get to them quickly enough – one of the incentives to build faster communications. 2
But there is another strategy, which is to use quantitatively sophisticated price prediction algorithms. The prediction doesn’t have to stretch too far into the future, a few nanoseconds will do. And this is where XTX has excelled. Rather than invest in faster communication equipment, it has invested in faster processing equipment. It currently runs 10,000 GPUs in its research unit and maintains 190 Petabytes of usable storage and 4 Petabytes of RAM (up from 60 Petabytes of storage and 1 Petabyte of RAM at the beginning of 2021). The firm combines various market signals into a price prediction using econometric methods, neural networks and modern deep learning architectures. Indeed, its name – XTX – derives from a mathematical operation used in regression analysis.
XTX established its foothold by coming into the market via an untapped asset class: foreign exchange. Foreign exchange differs from other markets in a few respects: it operates close to 24 hours a day, there is no clearinghouse and it is dominated by dealer banks. Electronic trading came to the market early, but banks retained control. In order to protect themselves from high-speed trading firms deploying unfavourable ‘taking’ strategies, banks introduced a “last look” mechanism to give them a short window of time to decide whether to accept the trade or not.
XTX entered the market on the price making side, using its technology to create a preferred venue for trading without the need for a “last look”. It was helped in the timing of its launch. The depegging of the Swiss Franc at the beginning of 2015 caused mayhem in currency markets. When some dealers stepped away, XTX maintained a market presence, bolstering its credibility among customers. It also benefited from the fallout from the forex probe, when banks were found to have colluded to manipulate exchange rates for their own financial gain in the years up to 2013.
The roll-out was a success. Within a few years, XTX had toppled banks like Citigroup to become a top 5 liquidity provider in foreign exchange. This year, it is number five overall, with a 7% share of the market.
The firm used its know-how in foreign exchange to enter other markets, notably equities where it has an 18% share in public markets in Europe. With many more individual end-customers of equities than foreign exchange, it is a deeper, more profitable market.
But it also pits the firm more squarely against the speed merchants. In a letter to the Securities and Exchange Commission in 2018, the co-CEOs of XTX wrote: “XTX Markets believes that the race for speed in trading has reached an inflection point where the marginal cost of gaining an edge over other market participants, now measured in microseconds and nanoseconds, is harming investors… Market makers providing liquidity need to price to the average of the toxicity of the order flow they interact with, and to the extent that they are being adversely selected by latency arbitrage strategies, they must widen their spreads to account for that possibility, which in turn increases the costs of trading for all investors…”
XTX uses its technology to guard against this risk to its business. But it also steers its strategic positioning so as to avoid conflict with ‘takers’. The firm is an active participant on the Aquis stock exchange in Europe and even has a 9.5% stake in the business. Unusually for a public exchange, Aquis has an outright ban on ‘taking’ by proprietary trading firms’ algorithms. XTX has also built up private bilateral trading relationships with known counterparties, becoming the largest such “systematic internaliser” in Europe. Rather than operating in public markets, these counterparties deal directly with XTX at better prices given the absence of “toxic” order flow in the relationship.
We’ve talked before in Net Interest about the power shift that has taken place in financial services – how investment banks lost out to exchanges and lending banks lost out to private credit funds. XTX is another reflection of that trend. In many respects it operates like a bank. Its former co-CEO was ex-Deutsche Bank and a new deputy CEO has recently been recruited from JPMorgan. The firm is also more willing to use its balance sheet to warehouse risk than others in the industry. At the end of 2020 (last available data) its UK regulated entity had capital resources of £233.6 million ($320 million). The firm’s average holding period across the most liquid currencies it trades is 5 minutes, and for emerging market currencies it is 16 minutes.
But first and foremost, XTX is a technology company which means it runs lean compared with banks. The firm employs just over 180 staff, which it houses in a pretty cool office. If volatility is going up, and the firm can continue to outsmart if not outrun other high-speed trading platforms, then XTX could be an interesting place to work.
I’m not sure I would recommend Flash Boys, which is unusual for a Michael Lewis book. But Trading at the Speed of Light by Donald MacKenzie is a good recent book on how ultrafast algorithms are transforming financial markets.
The three entities are XTX Markets Technologies Ltd, XTX Markets Trading Ltd and XTX Markets Ltd, all of which report in GBP. Aggregate revenues were £1.48 billion in 2021 and net profit was £667 million. In addition, the firm has subsidiaries overseas, for example its Paris-based subsidiary, XTX Markets SAS earned €16 million last year.
A paper published for the Bank of International Settlements shows that the average FTSE 100 symbol has 537 latency-arbitrage races per day. That is about one race per minute per symbol. In the modal race, the winner beats the first loser by just 5-10 microseconds with the average race being worth only around £2 each. The paper shows that the top 6 firms are disproportionately aggressive in races, taking about 80% of liquidity in races while providing about 42% of the liquidity that gets taken in races. On a net basis, much race activity consists of firms in the top 6 taking liquidity from market participants outside of the top 6. “This taking is especially concentrated in a subset of 4 of the top 6 firms who account for a disproportionate share of stale-quote sniping relative to liquidity provision.”