It's become conventional wisdom that the frantic trading in Internet stocks is driven almost entirely by retail investors. But anyone who thinks only retail investors are driving Internet stocks is mistaken. Consider the 600-plus institutional investors at the Hambrecht & Quist planet.wall.street conference last week.

There was a lot of buying power in those meeting rooms, with institutional types listening intently as public Internet companies revised their spin and private Internet companies -- many in registration for their IPOs -- debuted pitches that will help them sell anything from cars to dog food on the Web. As just one example of the lunacy,



delivered a satellite feed of the spin lady from the

Home Shopping Network

pitching shares of Lycos instead of cubic zirconium.

So institutions own these names, but because retail investors have left an imprint on how these shares trade, trading may never be the same.

Wall Street legend tells the tale of a sign on the

Goldman Sachs

trading floor that says:

If you can't find the other side of the trade, you are it.

What that means is that a Goldman trader should constantly weigh the opportunity and cost of putting up the firm's capital to make trades. Nickel commissions are OK, but real money can be made, or lost, using the house account. Yet when it comes to Internet stocks, that thinking is hopelessly out of date.

I learned this recently when my firm had the opportunity to sell a large chunk of an institutional-class Internet name. We used Hambrecht & Quist for the sale because it involved some

Securities and Exchange Commission

filings, so it was best to concentrate the shares with one firm, rather than shop around to disguise our size. Their traders crossed several hundred thousand shares in just two days. But what was most remarkable to me is that the average trade size was 200 shares. That's right, 200 shares.

At first I suspected that the buying interest came from

retail investors, from folks on




. It smelled like retail: As I watched the trades go by, maybe only five or 10 executions over the two days were for more than 1,000 shares.

But when I scrutinized the trading more closely, the action suggested quite the opposite: That it was institutions buying, because many of the bids were coming from Goldman Sachs and

BancBoston Robertson Stephens

. But no one -- neither H&Q nor Goldman nor any other market-maker -- would advertise more than 200 or 300 shares at a time on


Level II. What I finally figured out is that the electronic communications networks that match trades control so much of the flow that the big guys with big capital have to take tiny 200-share steps, just like the little guys.

Let me explain this first with a little Wall Street history, as I understand it.

Being a Nasdaq market-maker used to be a very lucrative business. But during the 1987 crash (remember that one?), Nasdaq market-makers stopped answering their phones because they knew it would be a sell order for a stock they were getting killed in.

Their punishment came when the SEC mandated a small-order execution system, SOES, so that anyone could execute up to a 1,000-share trade at any time on the bid or ask side. But this gave rise to what became known as SOES bandits, who began using the system to quickly buy or sell shares ahead of big customers when they saw a stock moving.

To counter these SOES traders, the market-makers widened trading spreads to protect themselves from these rogue traders. Depending on the stock, spreads of 25 cents between buyer and seller became quite common, and spreads of $1.50 a share were, and are still are, seen in the least liquid names.

Big spreads meant big profits, of course, which brought on the watchdogs. In a famous case, the SEC felt those profits were both obscene and somehow "proved" collusion between competing market-makers. The SEC fined most of the Street about $1 billion and made the brokerage firms advertise not only the main bid-ask spread that the world would see (the "outside spread"), but also the "inside spread," the price at which they might do the trade with another Nasdaq market-maker.

Once word of that spread was out, however, the ECNs came along and automated the process of quickly undercutting the inside spread to make money. Huh? Let me explain by example.

I'm trying to sell my Internet stock. H&Q offers some shares at a 32 1/4 ask, because I told them I wanted a 32 low.

Goldman has a 32 bid, probably because some firm wants to buy it with a 32 1/4 top. Now, if you got rid of the middlemen, I would love to sell my shares to Goldman for 32, but neither of these guys are in the charity business and wouldn't make money on the trade, so the standoff persists, 32 bid, 32 1/4 ask.

Like a real estate deal, these guys would probably be happy to meet somewhere in the middle by doing the trade between them at 32 1/8. They would give me my 32 and make the Goldman client pay 32 1/4, with each of them pocketing the 1/8th spread on either side: 12.5 cents per share, not bad.

But the trader on the ECN has different thoughts and wants to weasel in on this trade. So the ECN guy will shrink the spread down to 16ths. By buying at a 32 3/32 ask, and selling to a 32 5/32 bid, he can cross the shares between them, and keep the teenie -- the 1/16th, or 6.25 cents -- to himself. Even if this ECN trader had to pay the ECN 3 cents per share to execute the transaction, he made money. It's automatic because it is done electronically. He just programs his computer to trade between the inside spread when certain conditions arise, and away he trades. And his trading enhances liquidity, so everyone is happy.

Let's recap. There is a quarter spread, or 25 cents, which normally means each Nasdaq trader would normally get 12.5 cents per share. Now, some highly automated robotron trading system comes in and steals 6.25 cents of the spread, or more, leaving 18.75 cents or less to be split.

Of course, this is like explaining how a jet flies while it is on the ground. This stock doesn't stay at 32 bid-32 1/4 ask; it's moving like a yo-yo on steroids all that time. Volatility helps the ECN guy even more: He can hit the bid (in effect shorting the stock), move the stock down, have the ask follow the bid down and then cover by whacking the lower ask, all in a few seconds. Dangerous but effective.

Now, it is not in either H&Q or Goldman's best interest to advertise a large block of stock, thus showing their hand. And in the meantime, the automated ECN can recalculate the inside spread much faster than the human trader sitting at the Nasdaq terminal. Milliseconds mean megabucks lost, that yo-yo moves fast, hence all the liquidity in the stock is taking place at the 200- to 300-share trades executed on the ECN.

My point is a simple one: This stock market is built on liquidity. There is an assumption by institutions that they can get in an out of stocks in an orderly fashion. But the days of calling

Morgan Stanley Dean Witter

to say, "Get me out of 1 million shares of XYZ," and assuming the broker can get it done in a few phone calls? Those days are over.

Trading desks are not putting their capital at risk to the same extent as they did in the past -- especially in a thinly traded and volatile sector like the Internet stocks, which trade 200 shares at a time. So some guy with ECN access who thinks he can make a few pennies a share is whacking the market-makers.

This comes at a price, of course: I got my trade done, and there were as many buyers as sellers. But the day the buyers disappear -- scared off by a bad tech report or a meltdown in Brazil -- don't expect Wall Street to eat the trade for you. And that fast-moving ECN trader? He'll be long gone.

Andy Kessler is a partner at Velocity Capital and runs a technology and communications fund out of Palo Alto, Calif. This column is not meant as a solicitation for transactions; it is instead meant to provide insight into the methods of venture capital, technology and investing. At time of publication, neither Kessler nor his firm held any positions in the companies discussed in this column, although holdings can change at any time. Under no circumstances does the information in this column represent a recommendation to buy or sell stocks. While he cannot provide investment advice or recommendations, Kessler appreciates your feedback at