Wednesday, December 1, 2021

NASDAQ will move to the cloud, but high frequency traders will still be able to pay to be "closer" to the data

Here's Popular Mechanics from 2011:

Hibernian Atlantic is building an undersea fiberoptic cable that will stretch from New York to London. The purpose of this engineering feat? To allow computers to complete financial transactions 5 milliseconds faster than their competitors. Finance is now increasingly dominated by automated trading, and to a computer 5 milliseconds is an eternity.


Algorithms automatically execute sales and purchases based on triggers in financial data. Every trader has his or her own investment strategies, but the software often uses the same data. And as always in the world of trading, the first orders on the books are the first ones executed. With computers racing each other, a millisecond can place an order at the head of the line, before prices change as more algorithms place similar orders.

NPR from 2014:

MICHAEL LEWIS: Well, the most sensational building is the New York Stock Exchange building in Mahwah, New Jersey, which it's a fort. To get to it, you go through, you know, guards and German shepherds, and it feels guarded. And inside the building is the matching engine, the place where buyers and sellers meet on the New York Stock Exchange, not on the floor of the New York Stock Exchange.

There are no human beings - it's just signals going into a box, computer signals. The building is enormous, and the question is sort of why is so big if the exchange is now just a box instead of thousands of people shouting at each other. And it's enormous to house the equipment of all the high-frequency traders who want to be near the exchange so they can get information from the exchange more quickly than you or me.

GROSS: OK, when you say more quickly than you and me, it sounds like there's a messenger running from the exchange to the office of the broker, which is exactly not what's happening. How do you get the edge in terms of time, of getting it a nanosecond before the other people do?

LEWIS: It's a couple of millisecond edge they're usually looking at, and the way you do it is just - is to minimize the physical distance between you and the actual exchange

William Gibson in 2014:

WSJ in 2020 (via Slashdot):

Hollow-core fiber is the latest in a series of advances that fast traders have used to try to outrace their competition. A decade ago, a company called Spread Networks spent about $300 million to lay fiber-optic cable in a straight line from Chicago to New York, so traders could send data back and forth along the route in just 13 milliseconds, or thousandths of a second. Within a few years the link was superseded by microwave networks that reduced transmission times along the route to less than nine milliseconds. HFT firms have also used lasers to zip data between the data centers of the New York Stock Exchange and Nasdaq, and they have embedded their algorithms in superfast computer chips. Now, faced with the limits of physics and technology, traders are left fighting over nanoseconds.

Matt Levine today:

Nasdaq Inc. said that next year it plans to begin moving its North American markets to Amazon. com Inc.’s Amazon Web Services cloud-computing platform.

If you put the stock exchanges’ computers in the cloud, will traders still be able to pay extra for shorter wires? You’ll be relieved to know that the answer is yes:

Mr. Peterson said that in the first phase of the move, Nasdaq’s primary data center for its U.S. equities and options markets in Carteret, N.J., will be expanded and AWS will install computing resources there. Traders will be allowed to connect their servers to AWS servers the same way they currently connect to Nasdaq’s servers, he said.

You could imagine this eventually all becoming virtual; the traders could pay directly for lower latency. Instead of renting space for your computer next to the exchange’s computers at a data center in Carteret, just rent space for your virtual computer next to the exchange’s virtual computer in the cloud.