In fact, the companies at the tail end of the PC upgrade cycle - the risk-averse late adopters - will find themselves buying at the start of a new PC technology cycle. A raft of new things have been emerging over the past year or two, and they should hit the mainstream in 2005.
The most noticeable developments will be in processors. Pentium-compatible 64-bit extended chips are coming into widespread use, and dual-core processors should start to appear. These could have an impact in the server market, though there may be software licensing issues.
On the desktop, we can probably expect to see the PCI Express expansion bus replacing the now aged PCI and AGP graphics slots, and Serial ATA and Ultra ATA hard drives taking over from IDE. DDR memory chips (512MBytes minimum), USB 2 ports and Gigabit Ethernet should also become common.
However, Bluetooth short-range wireless, as a connector for keyboards, mice and printers, still looks too immature for widespread commercial use.
It would be nice to imagine IEEE 1394/Firewire ports becoming ubiquitous, but I suspect not. Few businesses seem interested in having them, let alone paying for them. Maybe the same thing goes for DVD drives too, though these are already standard in the home computer market.
Put all these developments together with the move from big glass tubes to flat LCD screens and 2005-2006 PC hardware could have little in common with what was popular in 2003-2004.
But of course, it ain't necessarily so. Motherboard manufacturers want to cater to the largest possible market, so "legacy" technologies have a habit of hanging around beyond the point where they became detrimental. The 1981-style Centronics printer port is one example.
On the other hand, motherboard and PC manufacturers are also coming under a different sort of pressure: for space. The 2005 PC is usually going to have a small motherboard and come in a smaller case, and there may not be enough room for redundant technologies.
Indeed, one of the most noticeable changes in the PC market over the past decade has been the shift of emphasis from desk-side PCs to desktop replacement notebooks. You cannot sell many tower systems nowadays, and this year we could see the mini-tower gradually giving way to cube-shaped boxes.
Today's users are less concerned about expandability, because so much stuff comes as standard, and they want their desktop real estate back.
Considering the cost of office space, it probably makes sense to give it to them.
Jack Schofield is computer editor of The Guardian