Something strange happened last week. Silicon Labs, one of the best engineering teams in the semis business, deliberately shrunk itself. Almost as strange, Skyworks, with one of the best operations teams in semis, finally managed to grow inorganically. Silicon Labs is selling half of itself to Skyworks.

The semis industry has been consolidating for over a decade. The industry has gone from something like 1,000 companies 20 years ago to 200 today. We have written about this in the past, speculating that there will only be five chip companies left in the not too distant future. There is a very strong trend for chip companies to get larger.

Silicon Labs and Skyworks both stood out from this trend, albeit in different ways. SiLabs has appeared on every M&A target deck presented to every larger company for years, but they have steadfastly resisted the trend, refusing to sell out. By contrast, Skyworks has been incapable of doing any acquisitions. They are notorious for being the low-interest, even lower bid in every sale process for as long as anyone can remember.

The more remarkable side of this transaction is Silicon Labs. They make a variety of products for a dozen different markets. These are mostly small chips that play some roll in larger systems. In semis, these kinds of devices are often subsumed into others’ products. But SiLabs has basically out-engineered the competition for a very long time, providing enough value to merit special treatment from systems designers and purchasing department cost-cutters. Their product portfolio has always been a hodgepodge covering wireless infrastructure parts, industrial systems, home networking and automotive.

Several years ago they began investing in the Internet of Things (IoT). As we have pointed out repeatedly, there is no such thing as IoT. It is really a loose term covering thousands of use cases across dozens of end markets. There is no “operating system” for IoT, no equivalent to the unifying power of HTTP/IP for the regular Internet. Silicon Labs set out to change that. They build a complex software ecosystem around their IoT chips. This is really hard for anyone today, and semis companies have historically been really bad at building software. It is important to remember that when we say “software” we do not mean some Android-like operating system (OS) or unified programming language like Javascript. Instead we are really talking about building device drivers and firmware that can run on all of SiLabs’ chips and support every combination of hardware and customer OS. This is closer to building a software interface that supports all of the world’s printers or graphics cards. (Perhaps the best comparison is with Nvidia’s Cuda library, a bet which turned out fairly well for them.) Messy work, that no one else really even attempted to build. Silicon Labs has done a credible job of it, but it has never been clear if this really had viable commercial prospects. Big investment with an uncertain outcome. Any other chip company would have given up a few hundred million dollars ago.

Imagine the set-up. A technically-focused chip company, investing heavily in a big complex project that has yet to deliver, in a market where they were probably fielding inbound M&A requests every Board meeting. And instead of caving into pressure and shutting down the IoT business or accepting any of the buyout offers this company sold their legacy and is now betting everything on IoT. This is a board with fortitude.

In exchange for selling the business that they have built over 30 years, they will get $2.7 billion from Skyworks, in cash. The company has committed to returning around $2 billion to shareholders, which leaves them with around $700 million which should be enough to sustain their IoT investments for some time. More importantly, this gives the company time to see results and breathing room from the activists who must have been circling, advocating for a sale.

Skyworks also makes out well. They are increasingly squeezed in their core market providing RF chips for mobile phones. With Qualcomm looming on the horizon, Skyworks also had to decide whether they were a buyer or a seller. Their problem has always been that the non-mobile side of their business, while nicely profitable, has never grown very fast. They have never been willing to commit to growing this side of the business through M&A, but now they have. The piece of Silicon Labs they are buying fits nicely with the rest of their non-mobile business, providing them the scale they need if (when) their RF business really starts to feel pressure from Qualcomm’s mobile RF business. We would not be surprised to see Skyworks someday do something similar to SiLabs and sell off their own legacy and focus on this industrial and automotive opportunity. Skyworks is now well positioned to become a scale player in automotive semis, one of the industry’s few high growth markets. This is not going to be good news for other auto-focused chip companies like Indie.

All in all, this is a deeply fascinating deal. Silicon Labs has built something that is unique in the chip business. If they can get it work and to scale it could be immensely valuable, and the company seems to think it really can succeed.