At a late-January meeting in a Marriott off the Washington beltway in Falls Church, Va., the Defense Department's main technology-research arm floated a proposal as nearly 100 scientists listened. They'd come from Boeing, IBM, Lockheed Martin, and other companies; from the Army, the Navy, and NASA; and from leading universities to hear a proposal for accelerating efforts to build a computer that theoretically could exist inside a coffee cup.

Within the next several months, the Defense Advanced Research Projects Agency, which will spend more than $2.8 billion this year on research and development for the Pentagon, is expected to launch a multimillion-dollar program to kick-start U.S. research in quantum computing, an esoteric area of inquiry under way at government labs, universities, and companies such as AT&T, Hewlett-Packard, IBM, and Microsoft. These supercomputers--built according to the strange laws of quantum physics, often operating at temperatures nearing absolute zero, and occupying spaces that can resemble a vial of liquid more than an electronic box--theoretically could perform within seconds calculations that take today's machines hours and solve in hours problems that might require centuries if run on state-of-the-art silicon.

If the research pans out, and there's no guarantee it will, quantum computers could revolutionize a computer industry whose main engine of economic growth--the doubling of computing power each year and a half predicted by Moore's Law--is in danger of losing steam.

The implications of quantum computers, which store and process information by exploiting the laws of quantum physics governing subatomic particles--"spooky action at a distance," Einstein observed--could turn upside down entire fields of research, including cryptography, engineering, weather prediction, space flight, and mathematics. Existing public key cryptographic systems would be useless. Timekeeping more precise than with atomic clocks could pinpoint satellites and spacecraft with stunning accuracy. Quantum computers could also yield exponential improvements in the speed of database searches and perform calculations virtually insolvable today. "Pratt & Whitney could use a 10-times performance improvement tomorrow if it were available at a reasonable cost," says Pete Bradley, an associate fellow for high-intensity computing at the United Technologies Corp. division that makes aircraft engines, in an E-mail message.

It won't be ready tomorrow. In fact, it could take 20 years or more to build a functional quantum computer. But consider the potential. The fundamental idea of quantum computing is that a particle--a quantum bit, or qubit--can represent both a 1 and 0 at the same time, so the number of calculations scales exponentially with each quantum-computing bit, compared with linearly in an electronic computer, where each bit must be either a 1 or 0. So a quantum computer using 14 calculating atoms, which is twice as large as what's been assembled today, could perform more simultaneous calculations (16,384) than the fastest supercomputer in the United States, at Los Alamos National Laboratory in New Mexico. (Los Alamos Lab itself spends about $11 million a year on quantum-computing research.) Scientists are a long way from those results, but the Defense Department program could pick up the pace of innovation with more funding. The U.S. government already backs quantum-computing research to the tune of $80 million to $90 million per year.

Darpa's proposed program, called Focused Quantum Systems, or Foqus, aims to build a quantum computer capable of factoring a 128-bit number--a common method of online encryption--in 30 seconds, with 99.99% accuracy. "Darpa has decided to put a huge chunk of money out for researchers to build a quantum computer," says Nabil Amer, the manager and strategist of the physics of information for IBM Research. "This will be a highly coordinated effort with the serious goal of bringing us to a go/no-go point: Will we be able to build this computer or not? Darpa all of a sudden got an epiphany."

IBM, MIT's Lincoln Laboratory, and other companies and universities are expected to participate in Foqus, with the goal of defining how to design a quantum computer, get information in and out of it, and correct inevitable errors. Darpa also is interested in the flip side of code-cracking: using quantum mechanics to create a new type of cryptography that couldn't be cracked with even the fastest factoring algorithms.