We have recently been diving deep into the world of analog semiconductors. There are a lot of interesting things happening in that market right now ranging from cyclical conditions and inventory levels, to automotive, to foundry, to all things China. And recently someone asked us if there was an equivalent to Arm in the space that might be an attractive way for investors to participate in the sector. This turns out to be a fairly challenging question to answer.
Arm creates intellectual property (IP) for processors. Put simply, they provide “blueprints” to chip designers of elements common to all processors. Imagine an architect is designing a house. They have to provide a complete set of blueprints for the whole house, but they will mostly get evaluated for how the house looks from the outside, how well the rooms flow together and how efficiently they make use of space. No one will notice how they design the house’s bathrooms. So instead of doing those time-intensive designs, they license someone else to provide the bathroom designs and then the architect copies and pastes those as needed around the house. Arm provides a similar service, albeit with something far more important and challenging than bathrooms. The use of these common designs saves the whole industry an immense amount of time and provides a layer upon which to build interoperable software.
Turning to the analog world, is there a company that provides something similar – a set of common building blocks usable across multiple chips and companies? The short answer is not exactly. Analog chips operate in very different ways than digital processors. They do not “make decisions” or use logic, they just respond to physical inputs. And while there are common blocks that can be used in multiple chips, these blocks remain a source of competitive differentiation between companies. There is a far higher degree of customization in the analog space, and having common IP across these chips would be counter productive.
The one area in which there is some commonality comes in the ‘hardening’ of each chip, transferring chip designs into actual instructions for physically manufacture. However, at this stage, this ‘hard IP’ is specific to each fab and their specific manufacturing process, and is already widely provided by the EDA companies and the foundries.
Put simply, there are no obvious sources of ‘soft’ IP in the analog segment, leaving little room for an Arm equivalent. We spoke to a lot of people in the industry searching for something like this, and found no obvious answers. We suspect there may be some lurking out there, and if you know of any, we would love to hear about them.
That being said, a recurring theme we heard was that analog companies lack digital capabilities. There is growing demand from customers for adding digital functionality to traditionally analog-only systems. For instance, one of the best-selling “AI” products in automotive today is a low-bit MCU with a small area of digital logic grafted on to it. Customers like this because it is cheaper and easier to implement than true processors from the digital chip companies. But as attractive as this product is right now, customers we spoke with are concerned that this chip has a limited future because the company providing it has limited digital design capabilities.
We think all of this leaves three areas of exploration:
- Is there a ‘soft’ IP company out there that could deploy across multiple analog products. Again, there are no signs of this, but we believe it may be possible.
- Companies which provide ‘drop-in’ digital solutions to analog companies. This is likely to be a promising field.
- Digital chip companies which are selling into heavily analog markets like automotive, industrial and aerospace.
The digital and analog chip worlds have operated largely independently for a long time, but the boundaries are shifting for many reasons, and that is likely to mean opportunity for someone.