The Federal Home Loan Mortgage Corp. has purchased a supercomputer, the first in the financial industry.
Freddie Mac installed the computer in April to reduce the time needed to analyze and price mortgage securities.
Supercomputers are powerful machines dedicated to complex mathematical or scientific tasks that would take other computers days or weeks to complete. The technology has been used in a variety of applications from calculating missile trajectories to forecasting the weather.
Prices Within Reach
A number of money-center banks have expressed interest in using supercomputers for trading operations, and several have experimented by renting time on them. But price tags of $5 million to $30 million have kept bankers from doing anything but window shopping.
That may now change. In recent months, prices on entry-level supercomputers have dropped dramatically, and many observers believe banks may follow Freddie Mac's lead and purchase supercomputers.
"It's clear that the entry-level systems have a certain appeal to Wall Street firms, and while I wouldn't expect an avalanche [of sales to banks], I wouldn't be surprised to see some trading-oriented banks buying in the next few years," said Gary Smaby, president of the Smaby Group, a computer consulting firm based in Minneapolis.
Time Saving Is Dramatic
Freddie Mac, which buys mortgages from lenders and packages them as securities, purchased the smallest supercomputer model manufactured by Cray Research Inc., which is based in Eagen, Minn. The purchase price: About $500,000.
According to Freddie Mac executives, the supercomputer will compute valuations at least 50 times faster than the personal computer-based system it is replacing.
This means that an analysis that took 10 hours can now be done in less than 15 minutes. In productivity gains alone, Freddie Mac estimates the Cray computer will pay for itself in less than a year.
However, while the productivity improvements from such a system are substantial, Freddie Mac executives warn that entry-level supercomputers are application-specific machines, not all-purpose computers designed to handle a wide range of functions.
Specialization Is Key
"It's important to note that we are getting away from general-purpose computing," said Michael Kenny, in Freddie Mac's Capital Markets division. "When you ask [the new system] to do a lot of things, performance drops dramatically."
According to experts, the need for a focused set of applications has led many financial institutions to limit their vision of where the supercomputers might fit into their operations.
Experts say trading rooms are the most obvious places to benefit from supercomputers' number-crunching and financial modeling abilities, but some say there are also applications for the machines to ferret out marketing opportunities from credit card account data held by many financial institutions.