The Necessary Evil

The software that controls semiconductor equipment is large, complex and difficult to develop and maintain. Millions of lines of code need to work in concert to meet requirements across some very diverse and challenging axes:

• High performance — Equipment control systems need to react in real time to carry out complex processes. Tool automation requirements for 300 mm have added a significant burden in this area. Many times, the software becomes the bottleneck in system performance. Software must be efficient and employ parallelism in the system to maximize performance.
• High reliability — Most fabs run 24/7. The uptime requirements for each tool are typically 98% or higher. To support this level of reliability, the software must be virtually error-free.
• Data intensive — With the complexity of manufacturing processes increasing dramatically as linewidths shrink, the amount of data that an equipment control system must process, store and transmit is large. The burden on the software is substantial.
• High flexibility — In general, tools have many possible hardware configurations and require almost infinite flexibility in the recipes that can be run. The software must be able to adapt to this flexibility while also trying to optimize the behavior of the tool.

The requirements are onerous and, unfortunately, this is just the tip of the iceberg. Many semiconductor equipment control systems are aging. The adding, patching, fixing and selling/acquiring cycles over the years have created many software systems that are virtually unable to be maintained. It takes more and more investment dollars every year to keep up these systems while new releases take longer and longer to get out the door and performance and reliability worsen, resulting in a tenuous situation. The investment to fix or replace one of these legacy systems is typically between $2–$15M. With chipmakers largely unwilling to pay for software and the equipment suppliers not having the money or resources to tackle the problem, the situation is only getting worse. Additionally, when a (brave) equipment company finally does decide to develop a new control system, most often the initiative stalls, is years late or fails outright. Offshoring of software development was thought to be a silver bullet, but most of these initiatives failed to deliver the expected results or failed outright as well.

The picture being painted is a negative one, but there are strategies that can be brought to bear to begin to address the endemic software problems:

• Look to software to add value and differentiate your products from your competition. Increasingly, the hardware aspects of semiconductor tools are becoming commoditized. Software remains a largely untapped source of potential competitive advantage.
• Develop a software roadmap that aligns with business goals and resource its initiatives. Ensure that your software architectures and technologies are aligned with the goals of the business for the future.
• Good software is expensive; bad software is vastly more expensive. Critical decisions regarding important software assets need to be understood and made by senior executives. Often, critical software architecture and technology decisions are made by the software guy with the strongest personality. Executives at equipment companies are not typically comfortable with software. Seek independent architecture and strategy advice. A small amount of money spent on an independent assessment from a qualified consulting firm can prevent a multi-million dollar mistake.
• Begin the process to migrate or replace legacy tool control systems that are not meeting the business goals of the company.
• If equipment companies do not value software, their customers certainly will not value it. Begin to charge for licenses, maintenance, upgrades, subscriptions, etc., even at a modest level, to begin to establish its value.
• Adapting to the cyclicality of the semiconductor industry is challenging. Staff your core software competencies and leverage development partners to staff the peaks. Identify and hold key software talent and software domain experts.

Begin to break the vicious cycle and turn the "necessary evil" of software into the "greater good" of competitive advantage, business value and return on investment.