In the last few days there occurred two events intended to shake the world of personal computing. National Semiconductor announced its intention to build a PC on a chip. Netscape hinted at its intention to bundle Communicator with Linux and throw the pair as a freebie competitor to Windows NT. One of these events is real and one is not, but neither will have the effect its authors intend because the market is already heading in a direction that can't be easily changed, if at all. The event that is real is National's PC-on-a-chip. I first wrote about this possibility back in 1992, suggesting the that it was six years away. Well National's estimate is more like seven years total, which puts my timing closer than usual to correct. Having developed, licensed, or acquired all the intellectual property to build the silicon bits of a personal computer -- CPU, graphics controller, peripheral controller, etc., National is putting all those bits together in an effort to beat Intel to a new product niche.

Intel has been concentrating on making microprocessors that are increasingly powerful, with ever greater numbers of transistors. National's approach is very different, since it requires from the start accepting the use of a less than top-of-the-line processor core. Simply integrating all these parts into a product that is ready for sale requires starting with well-proven parts, then there is the time needed for the integration, which also guarantees that the internal technologies are less than stellar. But this doesn't matter, because this is one specific instance when the final product is far better than the sum of its parts.

While Intel moves toward more complex chips, National is counting for its power on the integration, itself. Just shrinking the parts of a motherboard onto the CPU die has many benefits. The distance that electric signals have to travel is less, for one thing, which makes the product inherently faster. The size means it will consume less power, too. Building a single-chip PC may be more expensive to engineer, but the final product will be far cheaper to build since the actual manufacturing ought not to cost any more than building just the microprocessor in an Intel design. Final assembly is easier, too. And the ultimate result is a product that simply has to be more reliable, since there are so much fewer parts to break.

This is good stuff, this PC-on-a-chip, but is it going to result in yet another price point for personal computers? Are we going to see a $500 PC? A $200 PC? I don't think so. And Network Computers appear to be a joke.

The $995 personal computer is clearly with us to stay, but we're reaching the point where only so much of the cost can be wrung from the hardware. An unconscious conspiracy of Microsoft and its customers will keep the bottom end of the market from sliding much further.

There is already a very fine $500 PC in Brother's Geobook notebook, which runs the Geoworks operating system and comes bundled with most of the applications people normally use. But the Geobook just isn't tearing up the market. Bill Gates isn't losing any sleep over sales lost to Geoworks, because people who look at a Geobook but really want a PC tend to buy a PC. They can get such a good deal now for $995 that the $500 product is just too big a step down.

Consumers think of a personal computer as running some real version of Windows, and I don't mean Windows CE. But with Microsoft unwilling to discount Windows 95 under any circumstances, the price of a system with software can only go so low. Soon only Microsoft will be making any profit on these systems and that's simply because of the Windows monopoly.

Yes, Windows monopoly. You can't really call it anything else. What is it but a monopoly when there is such rigorous price uniformity and no discounting allowed? The price of applications has been dropping for at least the last seven years, the price of hardware has plummeted, but Windows and DOS haven't budged a nickel. That's monopoly. And since the monopoly shows no signs of going away, we can only suppose that $40-50 wholesale lump is going to eventually pin the price of the lowest PC.

What killed the application business was Microsoft Office. From Microsoft's perspective, the Office bundle has increased the total dollar amount it is extracting from each consumer for applications, but at a horrible cost to Microsoft competitors. I'm not complaining about this, mind you, because it has really helped consumers, but understanding the total effect is important to seeing where the software industry is headed. Microsoft now gets $300-400 from all of us for our first copy of Office and $150-200 for every subsequent upgrade. But go back to the pre-Office days and see how much it cost to buy that same functionality.

Half a dozen years ago, each major productivity application cost $300-400 and many cost $595. They came from different vendors and we tried to make WordPerfect work with Lotus 1-2-3, Harvard Graphics, and maybe Dbase. It was on the assumption that the world would continue to work that way and that software price points would remain where they were that Ray Noorda of Novell paid so much for WordPerfect and Quattro Pro. It wasn't the rise of Windows NT as a network operating system that hurt Novell so much as Microsoft Office turning Noorda's $1.1 billion application shopping spree into a white elephant worth only a tenth that amount when it was sold to Corel.

Microsoft Office is killing the applications business. The only way to survive is to be where Office isn't. Corel is reeling. Lotus SmartSuite is a joke. All that matters from Lotus, in fact, is Notes and Domino. The only strong players are those who live where Microsoft doesn't... yet. Symantec has utilities. Network Associates has anti-virus. Adobe has graphics. Microsoft has the rest. And Symantec and Network Associates thrive only because Microsoft hasn't yet gotten around to killing them. Only Adobe appears to be immune, because Microsoft learned a lesson trying to kill Adobe with TrueType back in the 1980s. When it comes to software technology, Bill Gates can't carry John Warnock's jock and Gates knows that.

Which brings us to Novell and this proposed bundling of Communicator and Linux, which Marc Andreessen has been discussing lately in public. It's a great idea, but it isn't going to hurt Microsoft. Where are the applications? If you count on WABI (Windows Application Binary Interface) to support the major productivity applications, then you are still giving business to Microsoft. If you are counting on a loose armada of Java applets, then you are two years too soon, and Netscape can't afford to wait those two years, while Microsoft can.

Netscape is in trouble. They need a new business model right away, and one that's based on giving their product away doesn't stand up to close scrutiny. What Andreessen is proposing is intended to scare Microsoft but it won't, simply because Microsoft's pockets are deeper than those of all the rest of the software industry put together. "Say Gates had to cut his prices in half," one Microsoft insider explained to me earlier this week. "At the new price point he could still go 3-4 years before he'd even have to tell Melinda, much less worry about the company. Can Netscape last that long?"

No.

Maybe Netscape is stalling, trying to think of a real strategy. Maybe they are trying to attract a suitor. More likely they don't know what they are doing. But whatever they do won't threaten Microsoft and it won't change the world.

In fact, there is no alternate operating system that threatens Microsoft. The only thing that does threaten Microsoft, and threatens it deeply, causing Gates to lose plenty of sleep, are devices that don't require an operating system at all.