Semiconductors lie at the heart of modern electronics. Without them, we would not have the smart devices, sensors, etc., that are becoming more and more integrated into our world. We are now seeing the rise of next-generation computing. This includes quantum computers, biologically-inspired computers, and nanocomputers. These new paradigms promise greater efficiency and security, but they require that semiconductors be designed to “keep up.” Innovators that can predict and win against semiconductor design challenges will win large shares of the market.
However, this comes at a time of growing difficulties. New technologies combined with the supply chain disruptions caused by COVID-19 and ongoing economic disruption related to both the pandemic and the war in Ukraine mean that there are design challenges now being faced by semiconductor manufacturers that are perhaps coming faster than they otherwise would have. Also, the U.S. has a talent shortage of designers. Add in ongoing pressure to make both manufacturing processes and the semiconductors themselves more energy-efficient, and we are in a difficult—but exciting—time for semiconductor manufacturers.
Top 5 Semiconductor Design Challenges
There are five primary semiconductor design challenges that manufacturers need to address moving forward.
1. Power Efficiency
Computers use energy. You have probably heard that mining bitcoin is bad for the environment, and it is because of the sheer amount of computing power it uses. Smart devices, of course, promise to reduce energy costs in other areas. But the fact remains that energy efficiency has to be a primary goal across all industries.
This means that manufacturers need to design semiconductors that use less power to get the same results.
2. Integration
Miniaturization is an important part of next-generation computing. We have already seen the same amount of power go from a room to a pocket. Increasing the power of small devices requires smaller chips.
This means semiconductors need to have a very small footprint to fit into this architecture. They also need to be integrated with minimal extra equipment.
3. Reliability
The growing trend of putting smart devices everywhere increases the risks if they fail. Internet-connected smart devices now control critical infrastructure such as the power grid. But hardware failures have recently caused issues in several areas. For example, in January 2023, a computer failure forced all flights in the United States to be grounded for a period of time. In this case, the issue was caused by human error, but it’s easy to see how a single chip failure in the wrong place could cause massive disruption and potentially endanger lives.
4. Cost
The rising prices of chips caused by supply chain shortages threaten the ability to keep devices affordable for businesses and consumers. Reducing costs is also a key part of remaining competitive—both for individual countries and for U.S. manufacturing as a whole.
5. Security
Returning to the issue of infrastructure, cybersecurity has become the number one priority. For example, ransomware often targets hospitals as they are more likely to pay up due to the fact that people can literally die.
A newer kind of ransomware, siegeware, attacks smart buildings by literally locking people out or by hacking systems so the building becomes uninhabitable.
Worries about cybersecurity attacks on power grids have spread across the world.
To make things more complicated, designers cannot focus on only one of these aspects but have to improve them all, ideally in synergy. So, what can designers do?
Solutions to Next-Generation Design Challenges
Solving these problems will likely dominate design trends over the next few years. Some of the responses to semiconductor design challenges that we are seeing are:
1. Power Efficiency
With current semiconductors, there is often a trade-off. Higher power efficiency equates to higher costs, often putting these devices beyond the reach of consumers.
Research on ways to make semiconductors more power-efficient is ongoing. For example, one team discovered that you could increase the switching efficiency of oxide semiconductor devices by inserting platinum nanoparticles. This technology will allow for the development of intelligent semiconductors. However, platinum is not cheap, and some experts fear we may run out. This means that we will also have to develop efficient ways to recycle platinum.
Other trends that are developing include ultrathin semiconductor films and 3D field-effect transistors. It is likely, however, that ultra-low-power semiconductors will remain expensive for some time, forcing a trade-off between cost and efficiency.
2. Integration
One developing trend is 3D chip tech. Major chip manufacturers are pursuing 3D layering as a way of increasing chip size past the current limit of photolithography tools. By “stacking” chips on top of each other, you can make thousands of connections that allow multiple chips to act as one. Currently, “2.5D” links in CPUs have the chips—or chiplets—right next to each other. But stacking is the way of the future now that we have developed techniques to reduce heat.
Most CPUs are now packages of chiplets rather than one large chip, allowing for innovative packaging techniques that are designed with the specific architecture needed. This innovative response to pressing semiconductor design challenges is growing in popularity.
3. Reliability
Chip reliability is not so much a design feature as a testing feature. It addresses the implementation side of semiconductor design challenges. Improved QA techniques are being developed to allow for advanced testing and validation, ensuring that chips with defects are not shipped. ATE validation systems can be used for both design validation and engineering validation and then display all data in one pass.
Modern data storage and analysis can help spot trends and thus quickly detect issues impacting the quality of produced chips, reducing the number of bad chips over time.
4. Cost
Any improvement comes with added costs, both the cost of design and the additional costs of new materials. The cost of new materials tends to come down with time and wider adoption. Alongside these costs, R&D funding is also often a challenge to obtain.
Reducing manufacturing costs is thus vital for ensuring new designs can come to production and keeping the cost of innovative devices reasonable.
Automation is one way to reduce costs. This involves investing in highly precise robotics. Delta-style robots with six-axis articulation can assemble circuit board packaging more efficiently than older-style “chip shooters.”
Real-time monitoring of QA, which we already mentioned, can reduce costs by helping fine-tune processes to reduce defective chips. Better power management is also part of the answer, which has the side effect of reducing the energy used in manufacturing chips. Design costs are being reduced by the use of advanced AI tools, which can also increase energy efficiency.
Finally, diversifying supply chains can also reduce costs by helping mitigate the impact of shortages and disruptions, such as geopolitical conflicts. For example, much chip production is in Taiwan and thus threatened by tensions between the U.S. and China.
Security
Traditional cybersecurity relies on software. But there is a move to secure the chips themselves. Hardware-level security is less likely to experience zero-day exploits and provides a baseline over which software security can be layered.
This is particularly important for smart devices, where much is built-in from the factory and where human error can easily result in a security hole that can have potentially disastrous results.
Case Studies
A few specific projects are showing promise in addressing some of these semiconductor design challenges.
Artificial Intelligence/Augmented Materials and Device Exploration
AI/AMDE is being developed by a collaborative team led by Professor Enrico Bellotti and Professor Luca Dal Negro and supported by a grant from the Army Research Office. They are developing AI-designed ultra-wide-bandgap semiconductor devices with extreme power density. These chips are designed for 5G devices, but they are hoping to expand further as the project continues. The team is working on both novel materials and new architectures.
Co-Design and Integration of Nano-Sensors on CMOS
Nanotechnology has a lot of promise for improving semiconductor technology. This project adds tiny light sensors to a complementary metal-oxide-semiconductor (CMOS) circuit. Rather than improving the semiconductor itself, they are working on adding these sensors, which increase functionality and can potentially be used for data processing.
Mimicking the Brain
Another research project uses biomimicry of the brain to develop semiconductors made of nanoscale artificial synapses and neurons. Our brain is a very energy-efficient computer, and copying it may allow the system to be more energy-efficient. It may also lead to semiconductors that are lighter and therefore use less material, potentially lowering costs and supply chain risks.
These projects will likely bear fruit over the next few years. It’s particularly worth paying attention to AI-assisted design, where supervised AIs can take on the task of fitting things onto the chip, improving efficiency by putting more functions on a chip.
Next-generation computing requires significant improvements in semiconductor design, and research is leading us in the right direction. In the interim, improved automation and next-generation materials will help face the current challenges. So will supply chain diversification which, while it requires significant short-term investment in new capacities, will improve the medium- to long-term efficiency of semiconductor manufacturing by reducing the disruption caused by a single event.
The semiconductor design challenges faced by manufacturers right now are:
- Simultaneously controlling costs while also increasing power efficiency
- Improving integration and miniaturization
- Dealing with vital issues of reliability and security that are only becoming more important with time
However, many of these can be addressed with, above all other improvements, novel materials, AI-assisted design, data-driven quality assurance, and new architectures such as 3D layering. The semiconductor industry is in a state of disruption right now, and it is important to make the investments in R&D that will carry forward into the future.
