Once upon a time, stock exchanges were packed with traders running, shouting, and elbowing one another on an open trading floor. Today, virtually all stock trading is done, well, virtually—through massive, globally interlinked computer systems. The rates of these transactions are now limited only by technology and, increasingly, by the speed of light. So a costly arms race has begun for telecommunications and network links that can give traders a competitive edge as small as a few tens of microseconds.
"Everyone is driving toward zero latency," says Graeme Burnett, who has worked for Deutsche Bank and ABN Amro Bank in the Netherlands and now runs Enhyper, a consultancy in England that specializes in financial engineering. "We've literally done every optimization you can imagine."
The impetus is a recent phenomenon called high-frequency trading. Typically, a high-frequency trading firm—or rather, its computer systems—buys and sells financial instruments while holding on to them for perhaps just fractions of a second. High-frequency traders make money by exploiting tiny and fleeting disequilibriums in the markets—say, when the price of one asset changes and the price of another that should be equivalent in value doesn't shift immediately to match.
The extent of such high-frequency trading activity in the United States—and to a lesser degree in Europe—is astonishing. (Because of certain taxes that East Asian governments impose on financial transactions, high-frequency trading hasn't made such inroads there.) X. Frank Zhang, a professor of accounting at the Yale School of Management, has researched the effects of high-frequency trading on U.S. markets and calculates that it "was responsible for about 78 [percent] of the dollar trading volume in 2009, up from near zero in 1995." He admits that his estimate may be a little high but says, "Everybody would agree that it's at least 70 percent."
High-frequency trading remains quite controversial. Zhang, for example, believes it leads to harmful price volatility. And high-frequency trading (or the sudden absence of such trading at a critical time) has been implicated in the "flash crash" of 6 May 2010, when stock prices in the United States plummeted by about 9 percent, only to recover most of that loss within 20 minutes.
One reason the high-frequency traders can beat others to the punch is that they often locate their computers in data centers run by the exchanges. The New York Stock Exchange, for example, moved its computer infrastructure last year to a facility it built in Mahwah, N.J., where it also leases space to trading firms. It can command premium prices for that space because close physical proximity means fast access to the exchange's raw trading data. Most other market participants don't see that data until milliseconds later, after it's been consolidated and combined with information from other exchanges.
High-frequency traders also gain an edge by having the fastest telecommunications link possible between distant trading centers where the prices of what is being bought and sold are fundamentally related. The fate of some companies whose stocks are being traded in New York City, for example, hinges on the price of commodities being traded in Chicago, and vice versa. If the computerized trading platform in one of these cities has access to information about the market in the other sooner than anyone else—even just a few milliseconds sooner—it can execute profitable trades.
That's why Spread Networks, of Ridgeland, Miss., invested what probably amounts to a few hundred million dollars last year to install a new fiber-optic cable along the shortest route it could find between New York and Chicago and then began marketing the connection to high-frequency trading firms. The round-trip travel time of a signal along its new cable, 13.3 milliseconds, is 3 ms faster than competitors can offer. With electronic trades now taking less than a millisecond to execute, firms with access to this fast connection can profit handsomely.
A similar effort is taking place to link New York City and London. In June, Hibernia Atlantic of Summit, N.J., began surveying the planned route for a new cable that promises to shave 5 ms from the 65-ms round-trip travel time of the digital signals presently being sent across the ocean. And Hibernia Atlantic, like Spread Networks, is pitching the advantages of its especially fast connection to the financial community.
Hibernia Atlantic appears to be thinking quite far ahead here. One of its advisers is Alexander Wissner-Gross, a research affiliate at the MIT Media Laboratory who (with Cameron E. Freer, a mathematician at the University of Hawaii at Manoa ) published an article last year in Physical Review E titled "Relativistic Statistical Arbitrage." That article investigates, in a rigorously quantitative way, what happens when the kinds of trading opportunities that high-frequency traders exploit are limited only by the speed of light, or at least the speed that light travels at in optical fiber.
"The naive solution is to put preprogrammed computers on either side of a low-latency link," says Wissner-Gross. That's what's happening now, and it's led to "an arms race to reduce point-to-point latencies," he says. But his research highlights a different strategy for making money. "The next phase is setting up at nodes." He means that the most advantageous position to be in, if you're trying to wring a profit from tiny discrepancies in price between two distant trading centers, is at an intermediate point between them. For example, if prices are fluctuating rapidly in New York but shifting only slowly in London, you should set up your trading machines between the two cities—but closer to the U.S. end of the connection. (Of course, this might put them, awkwardly, at the bottom of the Atlantic Ocean.)
The global financial system is indeed entering a strange new world. "You can no longer operate on the implicit assumption of absolute simultaneity," says Wissner-Gross, referring to the pre-Einstein notion of how time flows. So it seems finance is now moving into the relativistic realm. Perhaps that's why London's Financial News had to publish a clarification to a story it ran on 1 April, about a firm named Relativity Trading, which was reported to be able to execute trades "at the speed of light" and hoped one day to be able to execute a trade even before the decision to do so is made. It appears some readers didn't realize it was an April Fool's joke.
This article originally appeared in print as "Trading at the Speed of Light".
David Schneider is a Senior Editor at IEEE Spectrum. He worked for Scientific American and American Scientist before joining the Spectrum staff. In a former life he did research in geophysics. His beat focuses on computing, although he often blogs about drones and frequently writes for Spectrum's Hands On column, which betrays his MacGyverish tendencies. Schneider has, for example, build a synthetic-aperture radar using coffee cans and was able to detect an extrasolar planet using mostly stuff you can buy at the mall. He holds a bachelor's degree in geology from Yale, a master's in engineering from UC Berkeley, and a doctorate in geology from Columbia.