InformationWeek’s Charles Babcock saw NYSE Euronext CIO Steve Rubinow speaking at this past VMworld on how the NYSE’s cloud effort will democratize access to exchanges, and he drank the kool-aid. Babcock writes that the NYSE’s new cloud datacenter, which was built in a former rock quarry in upstate NY, will “open a privileged club” by giving (almost) direct access to the stock exchange to anyone and everyone.
But is this really true? Will the NYSE’s cloud platform really put a private wealth manager in Peoria, IL, on a level playing field with the Wall Street big boys? Certainly, many of the positive characteristics of cloud computing will be there: low-cost, on-demand, scalable, flexible, widely available access to a pool of remotely located compute resources. But like any other cloud offering, NYSE’s datacenter has internal tiers of service, and you have to pay up to move up. In this respect, the playing field won’t be at all level—it’ll just be massively un-level in a new and different way.
The main point I’d like to make in response to Babcock and the NYSE is that the NYSE datacenter merely swaps one exclusive club for another. That’s because regardless of what NYSE might want the public to think, low-latency access to its exchange computers is a scarce resource that goes to the highest bidder. Bidders who pay the most get to position their own servers the closest to the rack that actually executes the trades, so that they get the very lowest latency possible. The next highest bidder gets put further away, and so on. (Again, in this respect the NYSE “cloud” is like any other; on EC2, for example, you pay more for faster and larger instances. With any cloud, you get more when you pay more.)
So what the NYSE is creating in upstate NY isn’t some kind of open market utopia, where the solo financial planner in the Midwest will be trading on a level playing field with Goldman, Sachs. Rather it’s a new kind of private club, where the membership is restricted to deep-pocketed firms that can afford a) premium rack space in the exchange’s datacenter, and b) the very, very expensive compute and networking infrastructure needed to exploit the latency advantage that they get from being milliseconds closer than the next guy to the NYSE machine that executes the trade. The members of this exclusive, low-latency club will have an extremely profitable edge over everyone else.
So if by “democratic access” you mean, “an exclusive club where anyone with deep enough pockets can buy their way in to the top ranksracks,” I suppose it’s more “democratic” than, say, the Daughters of the American Revolution. But the NYSE looks more like a plutocracy than a democracy to me, and the playing field inside the NYSE’s cloud—or any other public cloud, for that matter—is never level.
A final point: the mythical Midwestern financial planner won’t actually be buying exchange access directly from the NYSE. Rather, she’ll be buying it from an large intermediary who is a licensed NYSE member that has bid for exchange access. Different intermediaries will no doubt offer different access speeds to clients, tiered by price, and the biggest Wall Street firms that are NYSE members won’t need intermediaries. So, meet the new boss, a much larger, cloudified version of the old boss.
Some historical perspective: pros and cons of networked markets
As Felix Salmon and I described in our Wired magazine piece on the role of AI on Wall Street, the past two decades of electronic market innovation have been both a blessing and a curse for investors. Under the old, pre-electronic regime, investors big and small were routinely gouged by intermediaries in a million different ways. For instance, in a conversation that we had with Kaufmann’s Harold Bradley, Bradley recounted how floor traders could get valuable intelligence about market activity from looking at the number of orders stuffed in a specialist’s front pocket. There were a million other ways for market participants to extract a large amount of value from their mere physical location on the exchange, or from their position in a chain of intermediaries linked by telephone and fax.
The move to electronic markets eliminated a ton of these opportunities for middle-men to gouge customers—in market jargon, they massively reduced “trading frictions”—but in the process they’ve left the door open for different kinds of shenanigans. Without going into detail, (again, read the Wired piece, or some of my past Ars coverage on this), I’d say that in today’s market, instead of a few privileged intermediaries extracting very high margins (through legit and non-legit means) from every trade, you have a much larger (but still very finite) pool of intermediaries and peers extracting razor thin margins from every trade. But the trades are happening at such insane volumes that you can print tens or hundreds of millions a year just by scalping fractions of a penny on individual trades.
At any rate, I promised a “pros and cons” list in my heading above, so there it is. As trading frictions, in the form of spreads and intermediary fees, are reduced by the shift to networked markets, the following good things happen:
- Spreads go down, so markets are more liquid (at least, they are when things are going well… see below).
- Fees go down as the number and size of intermediaries involved in the execution of your trade goes down.
- Access to markets is democratized, so that the little fish can swim in the exact same pool as the big fish, even if the little fish swims considerably slower.
And the following bad things also happen:
- Volatility goes up, because firms can and do enter and exit positions on a hair trigger. And why not, if it costs so little to get in or out, and all that trading speed that you’ve paid for ensures that you’re going to get out or in ahead all those other suckers who are lower than you on the latency ladder?
- Liquidity dries up instantaneously, as so-called liquidity providers simply shut down trading when things go south. (Cf. the May 2010 flash crash.)
- The little fish, i.e. the “dumb money” who is now thrilled by his low feeds and spreads, can more quickly and efficiently get his face ripped off by the larger, faster, better-informed, better-connected sharks that he now swims in the same pool with.
- Complexity goes up, so that when you have something major like the May 2010 flash crash, it’s very hard to get any kind of handle on what could’ve caused it. And forget about trying to prevent it by making up more rules. The level of complexity and speed of modern markets makes them functionally ungovernable.
Back in the 90?s I saw a TV interview with famous oddsmaker Jimmy “The Greek” Snyder, where Jimmy was being asked about then nascent spread of gambling outside of Las Vegas. (This was when states were just beginning to legalize gambling and allow casinos.) Jimmy’s take was that this was a very bad development, because people should have to get on a plane and fly to Vegas in order to lose their money at a casino; there should be logistical and financial barriers between the average punter and the casino door. In other words Jimmy’s argument was that democratized, low-latency access to the casino was very bad for the public, because it just meant that a greater volume of people would get fleeced by the house.
And after everything I’ve learned about our high-speed, computer-driven markets, I gotta say: I agree with Jimmy The Greek. The cloud is no place for a casino, and a cloud casino is definitely no place for the trading public. Some things should be costly for the public to get into and out of.