Windows critic shown HP's door

Windows critic shown HP's door
By Garry Barker
Technology Editor
September 13 2002

If Microsoft fears anything, it is Linux, the free, open-source operating system that is challenging the dominance of Windows software, especially in network servers.

Linux has been adopted by Dell, IBM and Hewlett-Packard who offer it as a parallel solution to Windows, yet, according to those of darker mind who watch the convoluted politics of the IT industry, even the biggest companies use a long spoon when supping with Bill Gates.

About 10 days ago, Hewlett-Packard fired Bruce Perens, one of the most respected Linux experts in the US because, as he told reporters, "the thing I did that was most hazardous for HP is the Microsoft-baiting I tend to do". The parting was amicable, he said, but he was fired.

Mr Perens was hired by HP two years ago to be a senior strategist for open-source software. In the US he was described as "an evangelist and rabblerouser on behalf of a computing counter-culture that increasingly is moving into the mainstream".

Service-related companies don't have the potential for profitability that software companies do. They are highly labor-intensive; whereas, once software is developed, it continues to generate profits.

That's a canned noise that we hear a lot, but it isn't really true in either direction.

Software development is also labor-intensive. When a new software product is released, one of two things happens: it either falls on its butt, in which case the company is lucky to even recover its development expense, or it takes off, in which case the company jumps on a labor-intensive treadmill called "getting out the next release" which -- if things go well -- goes on forever. Any company that develops a hot software product and then just sits there milking the first release, will quickly find that they lose their franchise to a competitor.

On the services side, a similar thing happens, although the consulting companies would like us to pretend that it does not. Let's say Accenture -- to pick a name -- gets hired to assist Yoyodyne with entering the credit card processing field. They will do a bunch of labor-intensive stuff, in the course of which they will get smart on the subject of credit card processing. They will have a Rolodex when they are done, and a fat report in the drawer. They'll take that same stuff and sell it twenty more times... not exactly the same stuff, but close enough that something like "economies of scale" start to appear. Internally at least, it starts to look more like a product than a service, although they'd never admit that.

And as long as that high-end box combined with that proprietary advantage provide enough value (and reasonable compatibility), people will continue to buy it.

I think you missed the point there. He was arguing that a hardware vendor could gain competitive advantage by offering an operating system that other people did not have, in other words, their own proprietary OS. That issue was settled by the market... the appearance of multi-vendor operating systems like UNIX and DOS put all the proprietary machines out of business. The only one left with any significant usage today is OS/390. VMS, AOS, Aegis, PrimOS... all those things are dead, and so are the machines they ran on. I think VMS even got ported to the Intel architecture, but no one cared.

When the PC first came out, Digital offered a desktop machine running RSTS/E. I thought that thing would be a big success -- there was a huge installed base of RSTS users -- but the customers stayed away in droves. They saw the handwriting on the wall for all proprietary operating systems.

[Concerning Sun:] I doubt that it will work because they don't have nearly the economy of scale that Intel and AMD provide.

I think you are right that reality is about to catch up with them. The original idea behind SPARC was that they could use the simplicity of RISC to keep the design-side costs competitive with Intel, even though they had much lower volume. And they used big Japanese fabs to manufacture the chips, so they slid by on the operations side as well. But I think that game is coming to an end. You can only squeeze so much performance out of RISC by increasing the clock rate, and they've pretty well maxed that out. There's nowhere else for them to hide.

I can't see Sun going with Intel and playing head-to-head with the likes of Dell. It's not their game. Here's my bet: they switch to IBM's chip, and within six months after doing so, merge with, acquire, or sell out to, Apple. A combined Apple and Sun, both running UNIX flavors on Power PC chips, makes a lot of sense.

That's a canned noise that we hear a lot, but it isn't really true in either direction.

You would argue with me if I said the sky was blue or the Earth is round. I'm not surprised.

Software development is also labor-intensive.

But unlike service revenues, the revenues from software development are tied to total units shipped. The cost of developing the software is a fixed cost. Once you spend the money, the only costs that remain are manufacturing (dup'ing the disks) which, as you well know, are marginal. With services, my labor costs are directly proportional to the revenues brought in. One job, one labor expense. Rinse, wash, repeat. Consequently, I'll have to disagree with you. If I had to choose between a software development or a service model, I'd definitely choose the former. The potential for profitability is much greater, as I originally stated.

As for economies of scale/selling stuff they've already done, it doesn't even approach the economy of scale of a Microsoft or IBM or Oracle software sales. You know, I know it.

The cost of developing the software is a fixed cost. Once you spend the money, the only costs that remain are manufacturing

I see that you have a brick wall for a mind. No, the costs are not fixed. If you want to stay in the business, you have to keep spending on more development expense, and you have to do that forever. You do not get to develop Windows 1.0 and ship it forever without further expense. If you want to have what the accountants call a "going concern," you get on the treadmill, and your development expense for keeping the Windows franchise goes on forever (you hope). While the people are buying 1.0, you're working on 2.0, and so on. It never stops; you never get to just keep the money and call it "pure profit."

It is also not true, as I will try to say for the second time, that expenses are linear in a service business. They are not. Once you develop an expertise on the linear model, you can then sell the same expertise to others. You'll charge the same, but it won't take you as long and you'll make more money. It is as if you were "manufacturing" expertise. Making the first expert is expensive, but the second and subsequent experts come much more cheaply.

It is also not true that development businesses are inherently more profitable. The only fair way to measure profitability across industries is return on assets, and the big services outfits do pretty well on that measure. Accenture typically has an ROA in the mid-twenty per cents. That is not at all shabby. There are many recent periods in which IBM's Global Services Division was the most profitable part of the company.

The thing a development company can do is grow faster. If you get lucky with a product, you can always make them faster than you can sell them. Hiring service employees at too high a rate of speed is a disaster, as many have shown, so there is some 'speed governor' on a services business that there is not on a manufacturing company. But if they do ever get to the same size, the return on assets (assuming competent management in both) will be comparable. You can't really argue against this, because if it were not true, Adam Smith's invisible hand would steer capital away from services businesses, and we don't see that. In the long run, capital is going to seek out the highest returns on assets; that's what capitalism is. If a sector consistently underperforms on that measure, capitalism kills it by withdrawing funds from it. If there is one obvious trend in capitalist economies over the last 50 years, it is toward more investment in services businesses. That would never happen if you were right. Therefore, you are wrong.

No, the costs are not fixed. If you want to stay in the business, you have to keep spending on more development expense, and you have to do that forever.

You're not paying attention. The costs are fixed for a particular release. That is what I've been saying -- and you knew it. Once you've spent money on a particular release, it can generate revenue at a far faster rate than a service-related business.

Once you develop an expertise on the linear model, you can then sell the same expertise to others. You'll charge the same, but it won't take you as long and you'll make more money.

Let's try to put this in perspective. MS spends approx $500M in developing Office. It then turns around and makes $5B selling it. You can do the math. Since service-related businesses are limited in the number of parties that they engage as clients, the labor and return on investment are damn near linear by comparison.

It is also not true that development businesses are inherently more profitable. The only fair way to measure profitability across industries is return on assets, and the big services outfits do pretty well on that measure. Accenture typically has an ROA in the mid-twenty per cents. That is not at all shabby. There are many recent periods in which IBM's Global Services Division was the most profitable part of the company.

Maybe in the dotcom salad days, Nick. In those days, MS, IBM, Oracle, etc were pulling in returns of 33% to as high as 50%.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.