The Mac Mini for most people is a great computer to hook up to your large screen TV. Why go with an Apple TV that is limited in functionality as well as power, when you can have a full blown computer that is a lot more powerful?
There is a completely different class of users - that use the Mac Mini in an entirely different way - Data Centers have sprung up with racks and racks of Mac Mini units - and their requirements are entirely different to what the Home Theater enthusiast needs.
For instance, the Home Theater enthusiast considers the Mac Mini to be essentially a low power computer hooked up to the TV - a glorified Apple TV for most people. But if you are putting stacks of the Mac Mini in a data center, you need the Mac Mini to be a powerful computer.
For a long time, Server Consolidation was the buzz word for Administrators - Take really monstrously powerful hardware and create Virtual Computers on them - to which you can dynamically allocate processors, RAM, disk space, etc. This was considered as a way to reduce the cost of a server. However there has been another dynamic at work - because servers use expensive parts that are produced in much lower volumes than the low end computers, the price per unit of performance is a lot better for the lower end computers than for servers. Despite the fact that price per unit is higher in servers, the cost of rack space, power consumption, heat dissipation costs, etc. still forced most people to go in for Server Consolidation. The Mac Mini has turned all those equations in favor of the low end machine. It takes up much lesser space, uses much lesser power, and generates much lesser heat - in fact, Mac Mini does not even require a fan to dissipate heat. Because of this, most "Server" applications can be efficiently handled by the Mac Mini. Even high end high performance Server applications can be handled cheaper by a stack of Minis that distribute the tasks between them, than by massive powerful computers.
The requirements of these two communities are completely different. For instance, the Home Theater enthusiast needs a DVD player/recorder - but on a server it is useless to have DVD drives on each server machine! They would much rather have that space utilized for 3.5" HDD instead of the current 2.5" HDD.
There are other such differences as well - like the need for USB ports, firewire ports, WiFi, etc. Servers dont need these capabilities, but Home Theater users need lots of these.
To add to the above confusion, there is also the nascent Car Computer market - where the Mini is very popular. That market has entirely different needs to everyone else - starting from the voltage used by the computer, and the location of all the ports/controls!
Quite obviously Apple does not want to confuse the market by releasing different versions of the Mac Mini targetting different segments - this is so "un-Apple". Along with this, there is another dynamic at work - Apple does not earn too much of margins from the Mac Mini -and obviously, the Mac Mini is cannibalizing some sales from the iMac.
There is one way for Apple to handle both these issues - and that, is to sell the Mac Mini as a "component" based system, where the components fit together like Bricks. Each market segment would buy the components they need, and can ignore the components they dont need. Apple needs to come up with a connector technology that gives much better throughput between the components than is possible with existing technologies. For instance, servers would like to have a lot of HDD space - would be great if a Mac Mini can have 3-4 HDD "bricks" attached to it - and with connectivity that is close in speeds to internal connectivity. Quite obviously, USB/Firewire connectivity for disks will not be acceptable at all.
This connector technology would be proprietary to Apple, and they can earn money by licensing this technology to 3rd parties. It should be sophisticated enough that it would work for various purposes - fast data for HDD, etc., support USB, Firewire, Bluetooth, WiFi etc. for the consumer, support additional Consumer devices like Video Grabbers, Audio/Video amplifiers, etc.
This drastic redesign is what is causing so much of a delay in the release of the Mac Mini. If Apple can do this right, and can patent everything that goes into this, it will easily be able to capture significant chunk of the market for servers, as well as the entire Home Theater enthusiast market. At the same time, margins will not be impacted at all - because each of the components can actually be sold at even higher margins that Apple makes on their other product lines. And people will happily pay for the convenience and flexibility that this approach offers.
This will also be the best way to offer "true" upgrading capability to the market. Quite obviously, none of today's computers are upgraded often. Even computers like the Mac Pro, which can be upgraded, dont see too many upgrades in practise. However, if the components are all individual bricks, that fit together seamlessly, it will obviously be a much more upgradeable solution - the customer can add/replace a brick at any time.
All these images floating around on the internet are most likely photoshopped rumor posts. The other possibility is that Apple has not yet perfected the "brick" design, so there will be one more iteration of the Mac Mini, with marginal changes, before Apple goes in for the total overhaul described above.