Exponential growth in data density cannot continue forever. Sooner or later, some barrier to further progress will prove inelastic and immovable. But magnetic disk technology has not yet reached that plateau.
The impediment that most worries disk-drive builders is called the superparamagnetic limit. The underlying problem is that "permanent magnetism" isn't really permanent; thermal fluctuations can swap north and south poles. For a macroscopic magnet, such a spontaneous reversal is extremely improbable, but when bit cells get small enough that the energy in the magnetic field is comparable to the thermal energy of the atoms, stored information is quickly randomized.
The peril of superparamagnetism has threatened for decades—and repeatedly been averted. The straightforward remedy is to adopt magnetic materials of higher coercivity, meaning they are harder both to magnetize and to demagnetize. The tradeoff is the need for a beefier write head. The latest generation of drives exploits a subtler effect. The disk surface has two layers of ferromagnetic alloy separated by a thin film of the element ruthenium. In each bit cell, the domains above and below the ruthenium barrier are magnetized in opposite directions, an arrangement that enhances thermal stability. A ruthenium film just three atoms deep provides the antiferromagnetic coupling between the two domains. Ruthenium-laced disks now on the market have a data density of 34 gigabits per square inch. In laboratory demonstrations both IBM and Fujitsu have attained 100 gigabits per square inch, which should be adequate for total drive capacities of 400 gigabytes or more. Perhaps further refinements will put the terabyte milepost within reach.
When conventional disk technology finally tops out, several more-exotic alternatives await. A perennial candidate is called perpendicular recording. All present disks are written longitudinally, with bit cells lying in the plane of the disk; the hope is that bit cells perpendicular to the disk surface could be packed tighter. Another possibility is patterned media, where the bit cells are predefined as isolated magnetic domains in a nonmagnetic matrix. Other schemes propose thermally or optically assisted magnetic recording, or adapt the atomic-force microscope to store information at the scale of individual atoms.
There's no guarantee that any of these ideas will succeed, but predicting an abrupt halt to progress in disk technology seems even riskier than supposing that exponential growth will continue for another decade. Extrapolating the steep trend line of the past five years predicts a thousandfold increase in capacity by about 2012; in other words, today's 120-gigabyte drive becomes a 120-terabyte unit. If the annual growth rate falls back to 60 percent, the same factor-of-1,000 increase would take 15 years.