草榴社区 Cloud
Cloud native EDA tools & pre-optimized hardware platforms
Request a Free Trial →
Table of Contents
Some call it the AI revolution. Others call it the era of smart everything. And others proclaim the world is being digitized. No matter what label you prefer, the pace for the introduction of new and amazing products is breathtaking. And it’s accelerating. Every wonder how this happened? Is there a root cause for all this new, intelligent technology that surrounds us? It turns out the digital revolution was predictable. You can find the root cause by examining the supply chain for these incredible new products. Let’s take a look.
Who’s Driving What?
In a prior post, I explored what was driving worldwide innovation. In that piece, I concluded that “semiconductors drive worldwide innovation.” While that is technically accurate, it’s not the whole story. For many years, semiconductors, or chips, have driven market growth and even defined new markets. Companies like Intel, AMD, and IBM supplied the microprocessors that created the personal computer (PC) and a completely new chapter of consumer products.
Later (much later) companies like Qualcomm did the same thing to help create the smartphone era. I will resist the temptation to compare the processing power of a PC to a contemporary smartphone. We all know the spectacular progress made here. There are many more examples of chips making markets. They all have some consistent attributes:
- A chip architecture is identified that serves many products
- Big semiconductor companies put huge effort into making a chip that covers the market
- Product developers get the chip, documentation, and maybe a reference design
- A product is built around the chip, with a focus on differentiation
That last part is key. In the world of chips as a market maker, differentiation is achieved with industrial design and software, primarily. These are the degrees of freedom available since everyone is using the same, or a similar, chip. Application-specific standard product, or ASSP, is the name given to these kinds of chips. While software could drive differentiation in this model, the “creativity” of the software engineer was limited by the capabilities of the ASSP the software was running on. If a user experience was envisioned that demanded a response time of 100 milliseconds but the ASSP could only deliver 250 milliseconds, the idea was shelved.
The ASSP-centric paradigm survived and flourished for a long time, many decades actually. Thanks to the exponential improvements provided by Moore’s law, new generations of ASSPs kept pace with market demands. Big chip companies printed money, and end-product companies struggled to differentiate and gain market share.
There is another chapter in the story worth mentioning. Application-specific integrated circuits, or ASICs. These chips are built for one customer versus a market segment of many customers. They certainly cost more than an ASSP, since the development cost for ASSPs was amortized over many customers. But if software differentiation wasn’t enough, and the market was big enough, building an ASIC made sense.
ASICs were typically customized versions of the ASSPs they replaced. The product development paradigm was basically the same, with an added degree of freedom. Then, about 10 years ago, several forces began to take shape that changed everything.
What Changed?
Three events created a fundamental shift in product development, and the world is now different as a result. They are:
- Hyperconnectivity: Thanks to ubiquitous high-speed networks, everything is now connected.
- Cloud/edge computing: Massive compute capability is now available on demand. It feels more like a utility than a technology.
- Purpose-built chips: Advances beyond Moore’s law, such as multi-die integration and new architectures optimized for things like AI processing, have changed chip design. Chips can now be built that perform specific functions very well, functions that are defined by software.
The last item is critically important. The added degrees of freedom for chip design have now created the possibility that a chip can be built to optimize a particular software algorithm, essentially inverting the product development paradigm that has existed for so many years. AI algorithms are hungry for data. Thanks to the development of AI accelerators and access to ubiquitous sensors, these chips can now assemble massive amounts of data. And the network and cloud provide the needed infrastructure to do something meaningful with all that data.
In this new product development paradigm, software and not semiconductors drive worldwide innovation. The products created by this new paradigm are truly exciting. Software as the leader with chips as the supporting cast has many implications. These will be the topics of future posts. For now, I’ll leave with a few news clippings to illustrate the revolution that is occurring around us.
