The crystal ball sitting on my desk predicts that there will be more software in our future. And, it's no wonder. Trends in chip design are daunting—mask costs are exorbitant, while other development costs are rising exponentially as verification becomes more difficult and the cost of failure grows. Leakage currents are getting to the point where standard process technologies are prohibitively power hungry for consumer applications. Project teams are opting for custom process solutions that delay the move to new lithography.

Previously, the easiest way to get the next great thing had been to move to the next lithography level, adding more functionality at consistent cost or offering similar functionality at decreasing cost. This approach isn't working and it soon will be too expensive for most electronics companies as fewer chip designers find leading-edge technology affordable or practical.

Instead, programmable solutions are emerging as a practical answer. After all, they are standard products and all of the hard problems can be solved once, appealing to a diverse set of users. Programmability lets fabless semiconductor companies create single chips to address multiple markets, amortizing research and development over higher volumes. Programmability has the potential to provide end users with a predictable design flow that shortens the project cycle.

These two trends will come together in an explosion of new architectural types within the next few years. Architectural innovation will replace innovation by scaling, and the architecture of choice will be programmable. As an example, check out Sony's new reconfigurable audio processor. It has created a reconfigurable device that lowers power compared to conventional processor architectures, yet remains cost-effective when measured against a dedicated application specific integrated circuit (ASIC) solution.

Analog Device's Blackfin is another example of architectural innovation. It combines RISC and digital signal processor (DSP) features, not by putting one of each on the same chip, but redefining the architecture so that it can do both at the same time, saving power and die size.

These trends mean major disruptions in chip design and the electronic design automation (EDA) business over the next five years. The electronics industry has already seen fewer design starts and we'll soon see fewer leading-edge chip developments. Handel Jones, of International Business Strategies, forecasts that ASIC design starts will decline 44% between 2002 and 2007. These trends point to far fewer chip designers in the future and many more software programmers.

Which is how the crystal ball came to predict more software in our future. Of course, the move to more software creates opportunities because all of these new programmable architectures will need effective programming environments. This generally means compiling, or synthesizing from a domain-specific language or environment, to an efficient implementation format that takes advantage of whatever inherent parallelism has been built into the machine.

DSPs are a good illustration for they represent an early and successful instance of the trend toward domain-specific programmable hardware. In fact, it is now a $7.5B market, according to Forward Concepts, and DSPs are present in virtually any device or technology containing the adjective "digital."

Opportunities for spurring continued growth in DSP markets through the use of a domain-specific language are close at hand. MATLAB from The MathWorks is widely used in DSP development for its extensive exploration and visualization capabilities as well as its DSP-specific libraries. Moving from conceptualization in MATLAB to implementation in a Texas Instruments or Analog Devices processor would shorten development times and improve programmer productivity compared with today's design flows. Companies such as Catalytic are addressing this challenge in the DSP domain with what's being referred to as DSP Design Automation.

In the future, I predict that there will be a variety of interesting opportunities for innovative companies looking for a technological challenge. And, I'm not looking into a crystal ball.