As a thirty-something year veteran of the electronics and semiconductor industries, I’ve seen a few changes in my time. I remember the first DSPs; the start of HDLs like Verilog and VHDL replacing logic diagrams; the first examples of on-chip RAM (anyone remember bit-slice processors?). And I’ve seen extraordinary changes in technology, from TTL breadboards in my early years, through to integrating multi-core 64-bit CPU+GPU-based systems for mobile in the past decade. It’s been an amazing journey, but driven by digital integration and high volume specmanship – and will bear little resemblance to what’s coming next.
The semiconductor industry matures
We’ve been used to high-growth startups in the semi industry – it’s an integral part of the lifeblood, driven in large part by the emergence of cost-competitive foundries and the fabless semiconductor vendor model. But I think few expected the astonishing rate of consolidation over the past 4-5 years, collapsing the industry to five or six Tier1s thrashing out the market, and pretty much everyone not in the top 3 either being sold or already absorbed. VC money has dried up for most semi startups – though there have been some promising signs of them starting to flourish again in areas such as AI and smart sensors
That means the mid-range vendors – often those that drove the greatest innovation and provided new solutions to more niche markets – have all but disappeared. And that’s disastrous news for systems designers wanting to create new innovative products. If you have to rely on software to differentiate your hardware, there’s only so much you can do.
We are going to need some serious hardware innovation, as we enter the smart, connected everything era. Radios, sensors, processing, data conversion and secure data must be integrated together into ever cheaper, lower power solutions for the IOT (Internet of Things) era. This will be driven not by “my CPUs faster than yours and has more cores too”, but by integrating complete systems, from the real-world analog interfaces to processing locally using miniscule amounts of power. So watch out specmanship addicts – that could kill your business in this next cycle.
With every cloud…
I’m often accused of being an eternal optimist, but the result of every consolidation phase, regardless of industry, is new opportunities. Every time. One reason for this is the release of thousands of highly skilled people at all levels, eager to do something new, and not afraid to engage with what’s coming next. Another is that in the process of driving a new product category as disruptive as the smartphone, we learned an awful lot about how to take really advanced technology and get it into volume production – right first time – in astonishingly short timescales.
Indeed this collapse in design to production from 3-4 years to 6-9 months for something as complex as a mobile phone has ramifications for us all in the coming decade. Can we apply all these advances to much more specialized markets?
Technology cycles have always been and always will be
While the first 5-6 years of the smartphone revolution were tremendously exciting as we kept on adding new capabilities never thought possible in a mobile product, the past few years have been, frankly, increasingly boring. As with most markets, once we run out of truly new things that are more exciting and unexpected than the previous model, we descend rapidly into a price war and consolidation.
That’s usually good for forcing the technology platforms to become more mature, more efficient and better supported, but maturing markets work for the biggest players far better than the small ones. As soon as serious volume enters the equation, the smaller players need to either grow fast or move on.
For smartphones the past decade has been of high value kit in unbelievable volumes: 1-2bn mobile phones per year dwarfed the 200m-400m PC volumes. And whilst margins have collapsed for most, those volumes are plateauing, not dropping. So now we have a fantastic, ongoing engine-room continuing to drive innovation in OEM and semiconductor technologies. So what else do we do with it, when virtually no other market has a fraction of those volumes?
Learning from the last decade to benefit us all in the next
We’ve learned so many things in the past decade: from how to design absurdly complex SoCs (systems on chip) to optimizing chip to product timescales; using open source software on a global scale to produce sophisticated yet open and constantly evolving global platforms; making technology easy to use for billions of people that previously could barely turn on a PC. It’s been quite a decade.
The investment community would dearly love to find another “perfect storm” market like mobile – but almost certainly they’ll never find it. Their latest darling – automotive – is not only far lower volume ( less than 90m cars shipped world-wide in 2016), but also hopelessly fragmented by brand, region and demographic. And while AI is an amazing technology, it is being rapidly reduced to a new set of processing engines just as we did with graphics, communications, and other compute-rich technologies. And I don’t think many people are going to buy “an AI” as they do buy phones. The world is gonna change like never before.
The era of niche markets and customization
Historically we’ve mostly used high volume to drive disruption and innovation for the past few electronics technology cycles. However I believe the next will be driven by customization at much lower volumes for much more specialized applications. We’re already seeing this in the car industry, delivering a bewildering range of models and options at every price point. Most cars are now customized per individual on the production line. So let’s now apply similar thinking to the electronics industry to drive innovation in IOT, industrial automation, agritech and many other applications where society has a genuine appetite for innovation and change.
This has to start with the hardware platform – that’s where the cost must be minimized, and performance optimized. Regardless of whether it has a small MCU or the largest multi-core heterogeneous processor array, the hardware has to work at the lowest possible power consumption, and deliver exceptional performance. Most of these innovations won’t be about how much processing muscle they have, but how they integrate optimally with sensors, actuators, radio communications and power supplies – mostly analog functions.
Disruption will come in small packages
So I believe this next wave of innovation and disruption will be fueled by a new generation of mixed signal chips, highly customized and optimized for end applications. The biggest difference: thanks to what we learned designing chips for OEMS building mobile phones from chip to end product, mixed signal chip design expertise will become increasingly integrated into OEM’s end product specification and design teams. The next OEM winners will be those that find ways to design their own chips that deliver new levels of price competitiveness, functionality and performance demanded by users of their end products. And perhaps this time the semiconductor teams will not get so obsessed by esoteric low-level parameter one-upmanship, as has dominated too many mixed signal chip competitors in the past.
The analog design world has always been driven by specialists – and that won’t change, whether at chip, board or system level. Analog is something that can never be commoditized like digital – but it’s how the real world works and always will. Intelligently customized mixed signal SoCs – driven by analog performance in real systems, not GHz or GFLOPS – will offer the product disruption so eagerly sought by innovative OEM startups, OEMs and ODMs.
Those OEMs will be the ones creating real solutions solving society’s biggest challenges such as providing better energy distribution, growing food more efficiently and consistently, automating specialist factories, democratizing healthcare for ageing populations and more. It’s time we used our technology revolution to empower the smartest specialists around the world to deliver innovation to thousands of real world applications for real people.