Jeff Robertson – July 22, Final – 735 words. Posted Aug 6-7/14.
Communication theory gave us the concept of the signal to noise ratio as a measure of the useful information in a message. I propose a new metric: buzzwords per square meter. The highest values seem to be found at trade shows and industry conferences. Semicon West 2014, held in San Francisco in July, was no exception. It was hard to find a presentation which didn’t reference the Internet of Things (IoT). It’s easy to see why: with analysts sizing the market at $30 billion annually ($5 billion for semiconductors alone), everyone wants a piece. Or, as was the case at Semicon, they want to persuade you that you’ll get your piece if you use their product.
The premise of the IoT is that anything that could be connected to the Internet will be, whether it’s a wearable (a Bluetooth-enabled running shoe evaluating your stride), your thermostat, or even a mundane FedEx package. The only way to get to billions of connected devices, compared to merely millions of smartphones, is to find a way to make these devices much more cheaply. Luckily, they don’t have to be as powerful as a smartphone, so that helps.
Now, when you gather engineers at a conference, they naturally want to talk about the latest “hero” demonstration. In the optical transport business, it’s sending the most bits down the longest piece of optical fiber. With sports cars, it’s top speeds and zero-to-sixty times. At Semicon, it was all about who’s got the smallest semiconductor manufacturing process geometry, and who’s going to be first to the next one. But it’s at the older, larger geometries that the IoT is really going to happen. There are two main reasons for this.
First, devices made at 65 nanometers (nm), 45nm, or even 28nm are much less expensive. The yield problems being experienced at leading-edge sub-28nm nodes are not going away any time soon, and costs appear poised to stay high. (Moore’s law seems to fraying around the edges – we haven’t hit the physics wall yet, but the economic advantages of the die shrinks don’t seem to be as reliable as they were in every past transition.) Luckily, we don’t need the most powerful ARM processor in our Bluetooth sneaker; a simple 90nm microcontroller is sufficient even for the Samsung Gear Fit.
Second, the desire for compact, integrated designs motivates the use of embedded SRAM and Flash, and embedded SRAM isn’t scaling well below 28nm. Heroes may capture the spotlight, but there’s plenty of work left for the everyman to do.
On the chip packaging side, the future isn’t so easy to read. Engineers are a pragmatic bunch, and they’ve tried pretty much every packaging technology for mobile chips, particularly for wearable applications, and there’s no clear winner. Cost will again be the primary driver, but IoT applications can have some pretty demanding environmental requirements, so the cheapest solution may not be the right one.
Finally, we have sensors. Sensors are the raison d’etre for IoT devices, particularly medical/fitness wearables. A big part of the reason we’re able to even able to conceive of the IoT is that the cost of sensors has fallen through the floor in recent years, thanks to the enormously competitive environment of the smartphone market. That means that margins are pretty poor. When asked who was going to profit from the IoT, Kurt Peterson, speaking at the Silicon Innovation Forum, would only say that it wouldn’t be the sensor makers. But as they advertising saying goes, “How do we do it? Volume!” According to Gartner, the sensor market grew at 6% in 2013, and is expected to grow 10% annually through 2018, so those manufacturers who are able to scale and find innovative cost reduction opportunities will succeed.
As an analyst, it’s fun going to trade shows and trying to separate fact from spin. One thing that can’t be spun is that Semicon was smaller than in past years. This is likely due to ongoing consolidation, which is a sign of a mature industry. One cynic called consolidation the enemy of innovation– opining that big sclerotic companies favour the status quo – but I saw no shortage of innovating start-ups offering everything from new FinFET architectures to silicon nanowires. Every one of those start-ups made a point of emphasising their patent portfolios. Innovation is alive and well, and the innovators are relying on their patents to protect and reward them.