Earlier this month, a truck pulled up to Building 50, on the grounds of the Stanford Linear Accelerator Center at Stanford University, in Palo Alto, Calif., and delivered a data center. Prefabricated, encased in a shipping container, and crammed with 252 servers from Sun Microsystems, the Blackbox, as it's called, now sifts data for physicists examining the effects of subatomic particles colliding at nearly the speed of light.

Buying and beta testing the Blackbox, the new self-contained data center from Sun, was a way of adding nearly a third more computing power without the expense of new construction, said Randy Melen, the head of high-performance storage and computing for Stanford's department of Scientific Computing and Computing Services.

"We needed to expand quickly this fiscal year, but solving the cooling and power challenges for the building takes longer," Melen explained. "We worked with Sun to answer the question, 'How do you extend your data center without too much pain?'"

That question and Stanford's plight sum up the future of computing demand, according to Sun CTO Greg Papadopoulos. Demand from Web 2.0 companies like Google, YouTube and MySpace; major data-crunching industries like pharmaceuticals, financial services, and energy; and software-as-a-service providers like Salesforce.com exceed the ability of Moore's Law to keep up.

"We are making a really big bet that this is what the future looks like," Papadopoulos said in a February presentation at Sun's analyst day that has become widely broadcast over the Web. "Systems engineering and the ability to work with brutal efficiency at deep scale is going to be defining in the marketplace and is going to separate off who ultimately wins in this business."

In other words, what Sun VP for Web 2.0 Peder Ulander calls "the Google model" -- building out massive server farms using hundreds of low-end machines -- is going to give way to "utility computing," where CPUs are provided either over the Internet or in high-end, massively dense units like Blackbox.

"The natural tendency ... toward building a small number of very large data centers" is being counteracted by a variety of forces, including real estate costs, power consumption, heat generation, and inflexibility, wrote Microsoft Research's James Hamilton in a paper called "An Architecture For Modular Data Centers" earlier this year. "All diminish the appeal of the large, central data center model."

Every major transition in IT models needs a good buzzword, and Papadopoulos has come up with a doozy: "Redshift," referring to the tendency of light from galaxies moving away from Earth to move toward longer, and thus redder, wavelengths. Companies with insatiable computing demands are "redshifting," and will enjoy exponential growth in the coming years, Papadopoulos asserted; those with more conventional demands on IT infrastructure will grow at about the same rate as the GDP.

The redshift concept has been taken up by a variety of commentators, including Harvard's Nicholas Carr, author of "Does IT Matter?"

If Hamilton and Papadopoulos are right, Carr wrote on his blog Rough Type, "then the computer industry may, as I've suggested before, be going back to more of the mainframe model, where profits lie in high-end engineering rather than low-end assembly."

Of course, the Google model still has appeal, particularly for small online providers of rich content -- think MySpace a few years ago -- who have no choice but to smart small. For them, Ulander said Sun offers a pay-as-you-grow model that essentially provides storage and computing power as a metered service -- rather like electricity from utilities.

So far Sun's bet seems well placed. In the last year, following the appointment of Jonathan Schwartz, Sun's former chief software architect and a big Redshift proponent, as CEO in April 2006, the share price of the former Silicon Valley titan has risen by almost 40%.

And if your computing needs tend toward the supercomputer realm rather than the data center level, Sun has that covered as well: At the International Supercomputing Conference in Dresden, Germany, last month, Sun unveiled the Constellation, which is a highly efficient, low-power system based on interconnected Sun Blade 6000 servers -- in other words, a supercomputer-in-a-box. Designed to run complex applications, such as climate, weather, and ocean modeling, the first Constellation will go online at the Texas Advanced Computing Center at the University of Texas, in Austin, later this summer.