During the Laser Focus World Executive Forum, Accenture’s Miles Klingenberg gave the audience an insightful look at the semiconductor chip industry and helped draw a few parallels to photonics. He started his talk by discussing the industry’s complexity, where it’s headed, and some of the economic corrections it is about to go through (see video).
Much like photonics, the semiconductor industry continues to invest heavily in research and development opportunities. In fact, the semiconductor industry has the second largest R&D spend of any industry in the world at $71 billion in 2021.
And the investments go beyond the traditional players. Although a fab costs over $2 million, some of the industry’s biggest customers (Google, Apple, AWS) have started to create their own chips. “The relevance of this will be important as we move into AI,” he says. Likewise, VW has considered building its own fab, while Ford and others in the automotive industry have started to work much closer with their chip suppliers.
Influences abound
There are a lot of trends impacting the semiconductor space, including: onshoring production activities; AI’s movement from a buzzword to something with true potential; the growth of autonomous mobility, and the metaverse buzz with semiconductor companies now developing their own headsets/hardware. There is also the influence of quantum computing on the horizon, surfacing as a lever for tomorrow.
“We are also entering a time of correction. At this time last year, there was a tremendous amount of demand mixed with very little supply. Now we have little demand and oversupply as these things start to turn up,” he says.
Interestingly, venture capital firms are now starting to build or acquire semiconductor startups as well. “Historically, VC firms have never done that. Why? Because there’s a barrier to entry, and the learning curves are too difficult,” he says. “For semiconductors, the materials cost a ton to design and time to get a startup successfully off the ground in the semiconductor space has been nearly impossible.”
As such, the significance of VC firms taking an interest is substantial because of the applications where the market breakdown positions data processing and communications at the top. However, AI-based tools like the Google algorithm for natural language process known as ChatGPT is garnering significant attention.
Geopolitics continue to impact the supply chain, including barriers around semiconductor equipment sales to China, and efforts to onshore have companies focused on building resiliency programs to better accommodate future supply chain issues.
The talent shortage has also been very difficult to navigate, and it’s definitely hindered the semiconductor space. Adding to the talent scarcity, there’s a difficult entry barrier for a lot of these companies.
Emerging markets are sparking activity as well, with a growing number of semiconductor companies determining how they can enter the AR/VR/XR space, explains Klingenberg. “Metaverse is certainly dumping a tremendous amount of money and companies are determining where they fit either hardware or software,” he says. “The demand is also skyrocketing for electric vehicles, with companies jockeying for position trying to find the best ways to capitalize on the potential opportunities.”
Time for total enterprise transformation?
Many companies right now are wondering about new business models in order to compete in their landscapes. “While many of the leading businesses are big ships that take a lot of time to turn, the AI generation represents a time to embrace a reinvention business strategy—being nimble, fast, and able to transition easily to new operating models with digital competencies,” he says. “A lot of legacy companies have had this waterfall perspective. We put a ship in one direction and we sail that way for the next 20 miles. Well, you know what—there’s an iceberg.”
With an AI- and data science-fueled transformation, businesses can go to market faster with better designs, and dramatically reduce the cost of production and development, explains Klingenberg. “And we can create new experiences for our customers, suppliers, and essentially anyone involved in our entire value chain.”
So where should businesses focus?
Design and manufacturing are at an interesting impasse.
In design, traditional chip customers Google and NVIDIA are both using AI to develop semiconductor chips. Why is this important? Historically, the database in order to design processes to use has been linear in terms of transistors, wires, and logics.
However, last year, Google came up with a framework called Prime in which it was able to design the TPU AI accelerators using a type of deep learning framework that learns as it goes, known as reinforcement learning.
NVIDIA made huge strides as well with its AI-based design delivering 25% lower power consumption than human design with the classic EDA tools on the market.
“Some of the biggest chip providers in the game have relied on undercutting people with price to performance,” Klingenberg says. “It took a vast amount of machine learning training hours, but what both of these companies created is significant. It wasn’t any of the big players. And it took a software company to come in for this industry and design it themselves.”
This is a crucial moment for a mature industry to ask questions: Do we have the talent? Do we have the perspective? Do we have the know-how to understand what’s happening in the market today in terms of software, in terms of AI?
When looking at manufacturing, the question becomes, what are we trying to accomplish in manufacturing? It’s efficiency gains, lowering the time to ramp up production, and doing more with less in part because of the talent scarcity. As a result, many companies are designing frameworks in their factories they can only achieve through AI and data.
“The complex fab of the future requires on-premise data solutions with quality management, asset maintenance, predictive capabilities, and sensors that are going to be able to like to collect all this data,” Klingenberg says.
