It's not easy to thrive in the computer hardware business. It has massive R&D costs, huge chip fabrication or buy-in costs and significant raw material costs. The more stuff you build, the more factory space and machinery you need to invest in to build it.

Compare this with the software industry, which has few of these significant costs, and one wonders why companies continue to make equipment at all -particularly in today's market where competition is fierce. It's no surprise that companies like Palm and Psion have spun off their software divisions into separate, arms-length organisations. As the hardware market becomes harder to compete in, so the potential for the software guys to sell their wares as embedded systems - perhaps even to the hardware people's new competitors - grows. In some cases , notably Palm's, the two arms have continued doing their own thing with some success; others, notably Psion, have given up entirely on making kit and are concentrating on the software aspect of life.

So, as one of the best-known vendors that has stuck with the combined hardware/software manufacturing model, is it time that Sun started to consider its position in the hardware market, and redefine itself as a software company?

Well, first of all we should remember that, in fact, Sun did spin off its software arm some years ago into a separate company (SunSoft, Inc), albeit as a wholly-owned subsidiary of the parent. That said, SunSoft is still committed to making software for Sun kit, instead of - like Palm or Psion/Symbian - pursuing anyone who'll listen about the potential of embedding its software.

The main difference with Sun, though, is that it's more than a hardware and software company. If you bought a PDA from Psion, it came with some software and perhaps a warranty of some sort. Just like when you bought a CD player from Dixon's, you could perhaps add an extended warranty and a nice leather case, for another few quid, but that was about the limit to the things the company could sell you.

Now put yourself in Sun's position. Sales of equipment and the basic operating system are merely the tip of the iceberg compared to the added-value items you can bring to the customer. The value-add services with which you can tempt the customer are almost endless. First, there's additional software for managing the servers. Then there's the enhanced support contracts, with fast access to on-site engineers – after all, if a customer is buying Sun instead of some kind of off-the-shelf, Intel-based kit, they probably have a significant processing requirement that's likely to be business-critical. Add in some remote monitoring and fault-finding and there's some more revenue to be had – in fact, $2.8bn of Sun's revenue for 2003 came from the sale of support services. Then, of course, you have the training costs – to get the best from the equipment, the customer's technical staff are likely to need some level of training … oh, and Sun turned over just under $800m in training revenues last year.

Then there's the fact that competition in this arena is limited. If you're Palm or Psion, you'll have started to get nervous as soon as Nokia, Ericsson, Motorola, NEC, Sony, Blackberry and any number of others all started to make devices that did the stuff that yours does. If you're Sun, though, you can count the number of competitors on one hand – and the fact that you're in a high-value marketplace means that even if you lose some market share to the SGIs or the HPs of this world, your share of a huge marketplace is still a pretty big number. Because it's a high-value marketplace, the barriers to entry are high: we're unlikely to see a new manufacturer of mid-range or high-end equipment on the market soon, if ever, because it would be nigh on impossible to find significant investment to grab any market share at all from the current players.

The final thing to consider is that the Sun architecture, which is based around the SPARC processor, is still an attractive one. They've chucked away some of their proprietary interfaces in favour of standards (so S/Bus has given way to PCI, for instance), which has shut up some of the dissenters that have moaned they're too proprietary. Some of the hardware they make is standard and can interface happily to non-Sun kit; remember, this is a company that sold over $1.5bn worth of network storage products last year. And the bits that are largely unique to Sun – most notably the RISC-based SPARC processor – are still attractive, efficient and innovative even though we're now in the 21st century, and are in fact better advanced in some ways (specifically the move to 64-bit processing) than their competitors.

So no, Sun shouldn't stop doing hardware. Even if it was punting gear out of the door at cost, the chances are it would still be making money on the associated services. It's still a hard market to be in. There's competition from other platforms which, combined with the overall commoditisation of computer systems, and the industry-wide price cuts that have resulted, has brought down revenues for computer system sales from $12.4bn in 2001 to $6.2bn in 2003. But in a high-value marketplace, and as a company for which sales of value-add and add-on services can't happen unless you're selling something to bolt these items onto, Sun shouldn't be looking to get out of selling hardware just yet.