When I started in the PC industry 23 years ago, it was hard to argue that there even WAS an industry. Back then, if you wanted a computer to use at home, you pretty much had to build it yourself. State-of-the-art was the S-100 bus running at one or two megahertz, Zilog's Z-80 was the hot chip, and graphics pretty much depended on what kind of terminal — yes, terminal — you connected to your computer. The big names of the time were folks like George Morrow and Bill Godbout and George Tate — names that are generally unknown today. Morrow seemed to design nearly every computer that came out, including the one used by the IRS at his tax audit. (George fixed the auditor's machine, not that it gained him any slack.) Godbout's was for a time the largest PC company in the world. And Tate proved that regular folks could think up things to do with a database program (dBase-II).

One interesting characteristic all these pioneers shared was an innate ability to make money. Sure, George Morrow eventually went bust, but what killed Morrow Designs was George's inability — pretty much everyone's inability — to anticipate the impact of the IBM PC. But once we came to understand that impact and adjust for it, the money machine just kept on churning. The IBM PC made its competitors rich simply by expanding the overall size of the PC market.

Jumping to today, we still see profitable PC companies, and many of them have names that were familiar from those earlier days. Compaq was certainly there, as were Hewlett-Packard and IBM. Dell was around, but we called it PCs Limited. What we don't see, however, are a lot of PC companies today making bushels and bushels of money the way they seemed to back then.

Something has changed over the past few months, and suddenly it is hard to make good money in the PC business. At least that's the way it sounds if you read the business section. Part of this can be blamed on sheer scale. Gateway, which is the latest PC manufacturer to stumble, is bigger in its own right than the entire PC industry of 20 years ago. Poor Gateway expects a disappointing quarter in which it will have sold ONLY $2.5 billion worth of computers. Remember Compaq, Dell, and H-P are all bigger than Gateway, which is about the size of IBM in unit sales. We're not in Kansas (South Dakota, actually) anymore. This is an enormous business.

And it is a business in which it is increasingly difficult to make money. Even Steve Jobs is having trouble taking Apple to new heights of greatness or even revisiting some of the old heights. And the reason for this leveling-off of PC sales is clear to me. After 25 years, the PC market is finally becoming saturated. Bill Gates and Paul Allen set off a quarter of a century ago to "put a computer on every desk and in every home running Microsoft software," and they actually succeeded. With computers on so many desks and in so many homes, most of us have all the PCs we need.

So Intel has shipped some Pentium IVs, can you really tell the difference from a Pentium III? Only two groups of users really give a darn about the Pentium IV and they are gamers and Photoshop artists. There aren't enough Photoshop users to really drive the market and the gamers, well they've just about figured out that their graphics card is more important than their CPU when it comes to performance. The truth is that for almost all users the Pentium IV, or any other hot processor for that matter, doesn't really matter.

For a couple of decades, the business worked quite simply. There were always more people who didn't use PCs than did, so there were lots of potential new users to suck in. New hardware generations were introduced every 18 months, so there were lots of old users to upgrade. Serious users would tolerate being no more than one generation behind the development curve. So everyone who had a computer got a new one every three years, and there were always lots of new users to accept hand-me-downs or generally expand the market. Well, today the new generations come every six to nine months, and for awhile it looked like the "every second generation" algorithm would hold firm. But like Reaganomics, this was an artificial market stimulation that couldn't be sustained. Hardly anybody really wants a 1.5-GHz computer, you see, because we can't, for the most part, even imagine what we'd do with such a beast.

In all the parts of the computer system, there is always some aspect that seriously lags and holds us back. Sometimes it has been CPU power, memory performance, storage, or simple stagnation in software. Today, the CPU and disk subsystems are far beyond what we really need while network bandwidth is the limiting factor. For all the promise of DSL, most people don't have it. What people do have are V.90 modems, which we'll pretend can do 56 kilobits-per-second. Today's top microprocessor churns at 1.5 GHz. Three years ago, the modem standard was 33.6 kilobits per second and the top microprocessor was clocked at 300 MHz. Three years before that, the modem to beat was 28.8 kilobits and a fast machine was around 100-MHz. So network bandwidth is falling behind at exactly the same time that network performance is becoming crucial to our whole idea of personal computing.

The sad truth of computing is that, with very few exceptions, programmers won't write software until there is hardware available that can run that software. So here we are in a market pretty much defined by Internet performance and that performance is stalled. And directly as a result, so are PC sales.

It's not just that we all already have computers, but that we can't think of new things to do with the computers we have. Steve Jobs has been for months making these bold predictions that we'll all be making home movies with our computers, but I just don't buy that. What we do with home movies is shoot them then put them in a drawer or closet and forget they exist. It is an American tradition to do so. Yet Steve seems to think we'll shoot our movies, edit them into poignant little opuses about the state of modern American life, THEN put them in a drawer or closet and forget they exist. It won't happen.

With American business fully computerized and networked, the upgrade market there is limited. There are still some homes to go, but not as many as one might guess. And now there is competition from other devices that aren't computers. In Microsoft's emerging .NET strategy, for example, am I at a particular disadvantage if my platform is WebTV, rather than Windows 2000?

There is a LOT at stake here, more than most people realize. The PC industry represents more than $400 billion in sales. By the time we've thrown in all the software, networking, and ancillary products and services, this is a $1 TRILLION business. It is a business that has helped power our record economic expansion. It is a business that, if it turns sour, can drag America into recession. So maybe, just maybe, it's worth making a concerted effort to help. I'm not talking about subsidies, but sensibility.

I have a plan, you see.

We've faced this exact problem before, and the way out in the past was to change our software objectives. It is no coincidence that Windows is increasingly bloated with each generation because that bloat absorbs MIPs and encourages us to buy new PCs. Processor manufacturers are always out sowing money among software developers to encourage processor-intensive applications. And they'll continue to do this, though the market is against them. For the really big computing tasks, you see, we've figured how to gang computers. So if you really need supercomputer power, you can probably get it with the computers you and your friends already have. The other big MIP-burner at this point is probably real-time video compression, but that is bandwidth bound. If only we had better network performance, it might be worth getting a new PC to really let Grandma watch the kids on her PC. But we're not quite at that level of network performance yet.

So that's the answer — improving network performance as quickly as possible. Whatever it takes to get everyone the equivalent of a T-1 line to their home or office is going to drive the next level of desktop sales. And by not having that higher network performance, we risk stalling the industry and eroding the market capitalizations of the bigger players. Heck, we risk pushing America into recession.

The companies that have the most to lose are the processor makers and big PC vendors. Intel, AMD, Motorola, Compaq, Dell, HP, Gateway, IBM, and Apple probably have among them a TRILLION DOLLARS of market cap at risk over the next year. They should be doing everything they can think of to help DSL, cable modems, and high bandwidth satellite connections to succeed, but they aren't. Right now, the DSL companies are so cheap these vendors should just take them over. For $10 billion in investments and subsidies, they could probably stretch the current expansion for another three years, but I don't think it will happen. That's because these companies take their cues from Wall Street, and Wall Street wouldn't understand or instantly reward such behavior. They'd reward it two or three quarters later, when the effort would start to show results, but that's too late for Wall Street and probably too late for the rest of us, too.