Smart cities and devices, autonomous vehicles and smart medicine make it easy to imagine a world where there are chips in just about everything that defines our way of life. IHS Markit forecasts that by 2025, there will be more than 75 billion installed Internet of Things (IoT) devices worldwide. The semiconductor industry is already riding this wave and most industry experts see the same trajectory for the foreseeable future.
While emerging, state-of-the-art devices get a lot of attention, traditional integrated circuits continue to be the workhorse of the digital economy. This is great news for the semiconductor industry, as the capital expenditure (CapEx) requirements for increasing existing capacity are significantly less than those needed to build new capacity. Regardless, it’s abundantly clear that the world of “smart everything” is going to require some shifting of investment priorities.
Smart devices also present a host of unique considerations depending on the application. For example, if your smart refrigerator fails, it will cause an inconvenience, but the implications aren’t all that significant. If your autonomous vehicle fails, on the other hand, the consequences could be dire.
Rethinking Moore’s Law
For years, the semiconductor industry has been packing big processing power into increasingly smaller chips. As the traditional scaling approach to Moore’s Law has slowed (it’s more and more difficult to miniaturize a device), there has been pressure to find new ways to improve performance and manage costs through innovations in design, equipment and materials. With challenges like this on the horizon, it’s time for our industry to adjust its mindset and proactively address the key challenges that Fourth Industrial Revolution applications present.
To handle rapidly increasing volumes of data, more chips will be required—both advanced logic and mainstream nodes. While many “fabs” (semiconductor fabrication plants) can incrementally increase output through improved tool performance and yield, there is also the need to build new capacity in every segment. The market for memory is booming, but demand will likely continue to outpace planned capacity in the next few years.
The capacity challenge isn’t confined to the chip makers. Manufacturing a greater number of more sophisticated chips requires new and more precise equipment. In 2019, EUV lithography will become mainstream, and fabrication techniques like advanced dry etch, deposition and CMP are being introduced. More raw materials will be needed, especially wafers, but also etch gases, precursors, photoresists and CMP consumables.
Managing costs is a top challenge in our industry— adding capacity is costly and so is the innovation needed to improve device performance. For mainstream nodes, new solutions are needed that drive even greater efficiencies in the ongoing effort to reduce costs. On the leading edge, pushing the limits of Moore’s Law will require innovation at every step in the supply chain. As manufacturing processes become more complex, no single company will be able to provide solutions for everything.That means there must be a commitment to collaboration and a more concerted effort as an industry to solve these problems together.
Increased Process Complexity and the Importance of Reliability
The new materials and more sophisticated integration flows required to achieve higher performance devices also add fabrication steps and increase process complexity. While many industries have to do more with less, the semiconductor industry must do more with more — at least when it comes to complex manufacturing processes. Every new step added to our manufacturing processes creates a potential point of failure (or more) and solving each of these challenges is time-consuming and costly.
To address these challenges, the industry must continue to adopt new processes and materials to enable scaling, better contamination control and defect reduction to improve yield. Historically, cost and yield have been a big priority for manufacturers, but with the emerging IoT, there must be a better balance, with reliability having equal, if not more importance.
Many IoT applications won’t need to stress these new reliability requirements. Smartphones and many other consumer-facing connected devices are “good enough.” But when it comes to medical devices, self-driving cars and many industrial/infrastructure IoT applications, good enough won’t cut it. Design and manufacturing for reliability, in addition to performance and yield, will also require innovations in defect and contamination control. Long-term reliability improvement is not only a social responsibility; it will quickly become a competitive differentiator—especially for mission-critical devices—as the IoT grows.
The Industry’s Mandate: Closer Collaboration
Rather than addressing challenges step-by-step and vendor-by-vendor, companies in the semiconductor industry must closely collaborate to identify potential challenges and solutions in product roadmaps with input from across the supply chain. That means manufacturers, foundries, assembly and testing companies, equipment makers, materials providers, and component companies will need to work together to meet capacity, cost, performance, yield and reliability challenges. Improvements in product design, manufacturing, and contamination control are also needed. By working together, everyone can achieve higher standards, as well as lower costs and greater efficiencies.
Our industry as a whole has a vested interest in getting new products to market as quickly as possible. A reduction in—or even elimination of—friction points will be key to addressing the vast opportunities presented by the IoT age and help make the world of “smart everything” a reality.
Jim O’Neill is chief technology officer of Entegris.