One of the big themes in the current stock decline of telecommunications and high-capacity bandwidth companies is that they are overbuilding fiber capacity. As such, whereas last year the Internet and broadband companies were the toast of the market, some have seen their share prices drop by more than 90% as the huge capital expenditures start to weigh heavily on their income statements.

But a bandwidth glut? Pshaw. This is an argument that has come up ever since Qwest (NYSE: Q) started its bold plan to lay 25,000 miles of OC-48 capacity (2.4 gigabit) fiber along railroad tracks all over the U.S. The argument was that this fiber was so far in excess of what usage needs would be that Qwest would never be able to recoup its investment, so most of its network would lay fallow for decades to come. Then in 1998 and 1999, when the real commercialization of the Internet was taking place, these arguments quieted down as people and companies suffered delays due to installed capacity constraints, and the Internet was being increasingly viewed as a world-conquering technology.

Funny how a year can change things. Internet commerce and content companies are failing by the score, and telecommunications companies have dropped severely due to questions about pricing, commoditization, and fears of huge capital outlays. Some of this is warranted; most is not. Lost in the shuffle is the fact that more and more businesses are connecting their dispersed operations over this bandwidth, and the Moore-like exponential growth continues unabated.

There are issues, however, and there is some commodity element to bandwidth, but it is dependent upon timing and location. Even now that the long-haul capacity between Washington and New York, or in Global Crossing's (NYSE: GX) backyard, New York to London, has tripled in the last three years, new bandwidth is subscribed very quickly. When a new route opens, or when the termination equipment has been upgraded to allow greater throughput on existing facilities, there is a temporary loosening of the market, but it doesn't last. And anyone who tried to get onto one of the news websites on Election Night know full well that our network is finite. (Why and how is a different matter.)

Traffic experts have a behavioral theory known as "rubberbanding." Have you ever been in a traffic jam that seems to have happened for no other reason whatsoever than the fact that the road was crowded? One driver somewhere tapped on his brakes, causing the cars behind him to overreact, and "voilą!" -- traffic jam. Eventually the cars gain a more stable speed and space farther apart.

It seems that there are many cases of "rubberbanding" in bandwidth. One example is capacity projections, which sprang forward when the overall market was kind, but constricted into a tight ball now that there are questions about capital spending by the carriers. There is a usage rubberband, described earlier, where new capacity continues to cause buildouts on many of the large routes. There is a pricing rubberband, as newer capacity and greater efficiency causes formerly astronomical pricing levels to drop (this is VERY different from being commoditized).

These fluctuations will continue. But the trend for bandwidth requirements is and will remain up for the considerable future, primarily due to explosive growth of use by companies of network applications and increased dependence on wide area networks (WANs) and other distributed networks for data storage, benefiting such companies as Critical Path (Nasdaq: CPTH), Akamai (Nasdaq: AKAM), and even storage giants EMC (NYSE: EMC) and Brocade (NYSE: BRCD). Just as computer applications have the tendency to grow in size as the increasing memory and processor speeds increase, the ways we all use the Internet will continue to fill the capacity available.

The question is whether or not the economics of building data networks is changing. Clearly prices have dropped on a cost-per-bit basis for most networks, so the enormous, wanton capital expenditures of the past will necessarily be informed by companies' overall health. The biggest area of backlog is still on the local loops, which are largely the dominion of the Baby Bells: Qwest, SBC (NYSE: SBC), BellSouth (NYSE: BLS) and Verizon (NYSE: VZ).

But these companies (less so for Qwest), along with the competitive local exchange carriers, have put hundreds of billions of dollars into voice networks that they are now trying to convert to data. It will work, eventually, but at a much larger cost than that of the dedicated data network companies such as Level 3 (Nasdaq: LVLT), Montana Power (NYSE: MTP), and Enron (NYSE: ENE). Those capital costs are crucial.

Those smaller companies that have already assumed enormous debt, particularly PSINet (Nasdaq: PSIX), in order to build out state-of-the-art data centers are sitting on goldmines. The question is whether the bandwidth market realities will converge quickly enough for the company to be able to benefit from its efforts, or whether the risk the company took in buying and building so aggressively will knock it out of the game before it gets the chance. There are scores of other small companies, among them Northpoint (Nasdaq: NPNT), Teligent (Nasdaq: TGNT), and Winstar (Nasdaq: WCII), which are in the same boat: their usage capacity continues to skyrocket, but there is a real threat that it won't come fast enough or at a price point that allows the companies to show positive operating cash flow and service their debts.

But these things are quite different from a "fiber glut." It's not going to happen, and some cash-rich companies out there are primed to take advantage of the continued growth.