Company of the Year: Nasdaq

Under CEO Bob Greifeld, NASDAQ OMX plays the stock trading game better than anybody.

The market has been open less than two hours and already 900 million or so shares worth $25 billion have changed hands. In a given second the total jumps by $3 million to $5 million--all without a sound. Unlike the New York Stock Exchange a few blocks away, this exchange has no shouting traders, no crumpled trade tickets on the floor. At the Nasdaq OMX Group, a single technician sits in front of eight flat-panel computer screens in a quiet operations center, 51 stories above the World Trade Center construction site. On one screen, quotes blink on and off at speeds barely visible to the human eye. On another, a fever chart showing orders and completed trades scrolls along like the electroencephalogram of an agitated 2-year-old.

To the extent that the Nasdaq market exists anywhere, it's within a single rack-mounted Dell server in a rented data center somewhere across the Hudson River. That machine routinely processes 70,000 orders, cancelations and trades per second but can handle up to 250,000 per second--enough to deal with trades on the Nasdaq plus the London and Paris stock exchanges with room to spare.

An entire trading floor crammed into a suitcase-size computer: That's the future of exchanges, and Nasdaq was there first, having been all-electronic--floorless, that is--since its inception in 1971. In the early days the trades were by telephone; since 1983 they have consisted of computer clicks.

With roughly 33% of the total volume in U.S. equities, and 2,500 employees, Nasdaq OMX is rushing to push more stock trades as well as futures, options and other derivatives onto its superfast, supercheap servers before competitors like NYSE Euronext catch up. "As you add scale, your incremental cost goes to zero," says Robert Greifeld, 51, a former computer salesman who took over at Nasdaq in 2003 as it was being spun out of the old National Association of Securities Dealers, now the Financial Industry Regulatory Authority. "Our goal is to add more incremental trades at zero cost."

In a year of spectacular market meltdowns, Nasdaq OMX Group has capitalized on the turmoil. It is our Company of the Year.

The chaos in financial markets--to say nothing of exploding volatility--has been a windfall for exchange operators. Combined U.S. trading volume on all exchanges averages 10.6 billion shares a day, compared with 4.2 billion two years ago and 1.5 billion a decade ago. The recent increase in volume is accompanied by an explosion in volatility: The CBOE Nasdaq Volatility Index, reflecting short-term expectations of volatility in the Nasdaq 100 Index, surged to 80 from 20 or so between mid-2006 and October of last year. At four-hundredths of a penny per share, Nasdaq takes in $800,000 in fees on a 2-billion-share day, just for pushing electrons through its servers.

But there's more competition for that traffic. A 2007 federal regulation ordered brokers to route their trades to the cheapest exchange, not the one that is most convenient. In Kansas City, Mo., Bats Exchange, a three-year-old competitor, now handles approximately 12% of U.S. volume, including 12% of the trading in Nasdaq-listed shares.

Traders are also doing 7% of their volume in "dark pools," the electronic equivalent of a back alley where buyers and sellers transact anonymously, according to Tabb Group, a Westborough, Mass. market researcher. "People used to talk about each stock having a principal exchange," says Daniel Mathisson, managing director in charge of a Credit Suisse division that uses computers to direct trades to the lowest-cost exchange at any given moment. "Now the trading's going all over the place, and there is nothing to stop that trend."

So Greifeld plays offense, using cheap technology to get business. In 2005 he paid $935 million for Instinet Group, one of the largest electronic exchange operators, chiefly to get his hands on the Island trading engine, particularly fast and inexpensive technology developed by a young Brooklyn, N.Y. entrepreneur in the mid-1990s. Within months Greifeld scrapped Nasdaq's expensive Tandem computers in a Connecticut data center and moved Nasdaq to off-the-shelf servers. "We have to have the same cost structure as the startups--we can't give any quarter," he says.

Worker productivity slowed in the summer while wage pressures increased, but both developments were better than expected and are unlikely to raise inflation alarms at the Federal Reserve.

The Labor Department reported Wednesday that productivity, the key ingredient for rising living standards, rose at an annual rate of 1.3 percent in the July-September quarter. That's down from the 3.6 percent growth rate in the second quarter, but slightly higher than the 1.1 percent increase initially reported a month ago and better than the 0.9 percent rise economists expected.

Wage pressures, as measured by unit labor costs, rose at an annual rate of 2.8 percent, after having declined at a 2.6 percent rate in the second quarter. The rate of increase in the third quarter was the biggest jump since a 4.5 percent rate in the fourth quarter of last year, but was below the 3.6 percent advance originally reported and that economists expected.

The Fed closely monitors developments in productivity and wages to see if inflation is getting out of hand. But the central bank was likely to view the recent developments as temporary and not long-run trends.

Analysts had expected a big downward revision in productivity given the fact that overall output, as measured by the gross domestic product, was revised to show a decline of 0.5 percent at an annual rate, a bigger drop than the 0.3 percent decrease that was originally reported.

Still, the 1.3 percent rise in productivity was the weakest showing since a 0.8 percent rise in the fourth quarter of 2007.

While rising wages and benefits are good for workers, if those gains outstrip increases in productivity it can create serious inflation problems as businesses are forced to boost the cost of their products to cover the higher wage demands.

A small biotech company thinks it can make trading desks better by eliminating the people.

Colin Hill

In the midst of the biggest financial crisis in a generation, a tiny biotech company, called Gene Network Sciences, of Cambridge, Mass., thinks it can make Wall Street smarter. How? Get rid of the humans.

Their idea: Take the supercomputers Gene Network Sciences already uses to help Pfizer (nyse: PFE - news - people ) and Biogen Idec (nasdaq: BIIB - news - people ) invent drugs and use them to help hedge funds trade stocks, bonds,and other assets. "Computers and data are smarter than people," says Colin Hill, the theoretical physicist who founded GNS and will be chairman of its new trading spinoff, Fina Technologies.

"We believe the economy and the financial system are governed by complex networks, just like the genes that control cells and the neurons that control brains," says Hill. "And we believe that, using artificial intelligence, we can start to extract that circuitry from raw data."

Already, computers have replaced many of the guys who run trading desks. Now Fina wants to replace the computer-programming math whizzes with more computers. Perhaps one-tenth of fund managers are "quants"--short for quantitative traders--who run computer algorithms that buy and sell stocks so often they may account for one-third to one-half the trading volume on the New York Stock Exchange.

When it works, it makes fortunes. James Simon founded quant firm Renaissance Technologies in 1982, and is now worth $7.4 billion. David E. Shaw, a computational biologist, founded the quant firm D. E. Shaw in 1988, and is now worth $2.7 billion.

But sometimes, groups of these funds lose money at once--they are all using the same math, and thereby make the same bad bets. Fina thinks it can avert the problem by losing the phalanxes of mathematicians and their market-predicting equations and instead letting computers run the show.

Instead of trying to work out a way to predict the market using a human brain, Fina’s system would dump all available data into a computer system that takes snapshots of billions of possible predictive configurations. Imagine putting a picture puzzle together in that many different ways and deriving probabilities from all of them.