A chemical computer, also called reaction-diffusion computer, BZ compter or gooware computer is an unconventional computer based on a semi-solid chemical "soup" where data is represented by varying concentrations of chemicals. The computations are performed by naturally occurring chemical reactions. So far it is still in a very early experimental stage, but may have great potential for the computer industry.

Hey that sounds amazing ... can you give more details about the chemical computer?

Designing a chemical computer
a working prototype of a (bio-) chemical computer is being planned in many laboratories as a part of studying the information processing capabilities of bio-inspired chemical systems. A non-silicon parallel processor is the ideal biological computing substrate for studying and constructing non-classical computer. The silicon can be amorphous or fine grained. Instances of existing prototypes are:
- spatially distributed nonlinear precipitating systems
-excitable chemical systems
- tubulin microtubules(collision based and quantum computers)
-molecular automata and machines

Rationale
The simplicity of this technology is one of the main reasons why it in the future could turn into a serious competitor to machines based on conventional hardware.As compared to a conventional microprocessor, In a BZ(Belousov-Zhabotinsky) solution the waves are moving in all thinkable directions in all dimensions which enables it to handle billions of times more data than a traditional computer.

Basic principles
the BZ reaction has wave properties which means it has the ability to move information similar to waves. In the chemical computer, the logic gates are implemented by concentration waves blocking or amplifying each other in different ways.

There are four basic steps in producing a viable chemical, physical or biological computer.
1)Decide what the computer is going to compute: the right area of operation of the computer should be identified.
2)Choose a substrate that matches the problem: pattern recognition or signal analysis can be better done in neural networks whereas plane tessellation can be done in iquid-phase chemical media. Multiple parallelism should be made use of.
3)Interpret space-time states of the substrate in terms of information units:
Ways should be developed to encode data into initial disturbances of the substrate state and then decode results from the states of the substrate . parallel input and parallel outputs design would be advisable.
4)Subject the substrate to scoping inputs to reconstruct an input-to-output function realized by the chemical substrate:
a non-trivial interpretation should be provided of the space-time dynamics of the chemical/physical system.

For more details see;
http://en.wikipediawiki/Chemical_computer
ftp.ftp.cordis.europa.eu/pub/fp7/ict/docs/fet-proactive/chemit-02_en.pdf