50 years of Moore’s Law, and 50 years of technological advancements that have brought manufacturing profitability to semiconductors companies. Let's examines whether this same principle still holds true...
For more than five decades, companies have relied on technological advancements from Moore’s Law to drive manufacturing profitability. It was simple, yet brilliant, and it worked. Computers steadily became smaller and more powerful as transistors on integrated circuits became more efficient. Semiconductor companies reaped the rewards because their chips, when manufactured on more advanced technology nodes, actually cost less.
Now let’s fast-forward to 2020 and examine whether this same principle still holds true. The market dynamics have changed substantially over the course of 50 years, prompting concerns that doubling the number of transistors every one to two years might not be the right approach anymore.
As the industry has transitioned to newer nodes, design costs have grown exponentially. To put that in perspective, the annual design costs for 65-nm chips were well below US$50 million back in 2016.1 However, today, the design cost for a 5-nm device ranges from US$210 million to US$680 million, and for a 3-nm device the cost ranges from US$500 million to US$1.5 billion.2 That is a 3,000% increase as we go from 65 nm to 3 nm. With such skyrocketing costs, it now makes “Moore” sense for semiconductor companies to realign their long-term goals to identify additional opportunities for older, more economical technologies. While moving to advanced nodes may deliver growth in certain sectors, it will not necessarily deliver profit.
This current situation is in stark contrast to what semiconductor companies have been used to. In the past, they could justify high initial design investments because the cost of designing and manufacturing newer nodes came down over time as yield and production quality ramped up. Each successive generation of technology took advantage of lower manufacturing costs than the previous generation.3
Clearly, semiconductor companies will continue to leverage their vast expertise to innovate and drive new technological capabilities for a very long time. However, the decades-old mandate to double wafer capacity no longer carries the advantages it once did. As a result, chip companies need to pursue opportunities that are aligned to a viable business strategy as well as to their growth and profitability objectives. That means they need to make smart choices with their R&D wherein they really think about the costs, benefits, and requirements of moving to the more advanced nodes — and whether there is a market opportunity on which to capitalize.
Semiconductor companies should also make sure they are generating as much value as they can out of their existing investments in older nodes. Many of these older nodes are so established that they still have many years of life left to offer growth and profitability. Instead of jumping to the next node because that is what semiconductor companies always did, they need to step back and see if customers are even asking for it or will need it in the near future.
In summary, the end of Moore’s Law doesn’t mean that semiconductor companies will innovate any less. In fact, they may innovate more by doing less. This may not always mean an automatic jump to the newest, most advanced process node available. Instead, companies will look to new doors for innovation that they may not have thought of previously. The industry is going to be just as exciting, just in a different way than what was originally defined by Moore’s Law.
— Syed Alam is managing director for strategy and consulting strategy at Accenture.