Full disclosure: I am a MontaVista employee and thus a stakeholder in this acquisition. However, I have worked with embedded systems and Linux since 1992, including two years at Wind River in the mid 1990s. I hope I can present a somewhat-objective, partly-informed grasp of the facts beyond my current situation.
I was intrigued by the opportunities for MontaVista provided by Intel’s acquisition of Wind River earlier this year, and a little scared by Mentor Graphics’ swallowing of Embedded Alley, which nudged Mentor into the role of embedded Linux provider, albeit focused on custom solutions.
However, I am absolutely floored by the opportunities presented for both MontaVista and Cavium with this merger.
The background issue with this deal is that embedded systems are very diverse from a hardware standpoint, and there are many players in the field. There are base architecture providers/licensors that provide raw architecture flavors: ARM, MIPS, x86, PowerPC, m68k, SPARC, Alpha, PA-RISC, a whole host of them. The technology is often known as IP (“intellectual property“). From these spring actual hardware providers, chip and board manufacturers who license the actual platform technology and produce physical examples. These licensees come in many different modes:
- Some, like x86, are produced by the same company that invented it (Intel in this case) as well as by competitors (like AMD).
- Sometimes the IP is owned by one company and licensed to a number of manufacturers. ARM is like this (owned by ARM Limited), as is SPARC (owned by Sun, now by Oracle) and MIPS (owned by MIPS Technologies).
- Sometimes the IP is co-owned by several companies and licensed by a consortium (like PowerPC, developed by IBM and licensed by Power.org), which adds an interesting twist that consortium members can make money on their competitors’ products. It’s a crazy world.
- Sometimes the IP ends up being owned by a company that goes under, leaving it in a delicate state. A timely example is Transmeta‘s IP, which was owned by Novafora until they closed up shop a few months ago. The IP is now for sale. (This one is particularly interesting: the Transmeta Crusoe and Efficeon processors were RISC-based, but supported x86 instructions. Again, it’s a crazy world.)
(Please note that this doesn’t even qualify as a primer on the subject. I’ll try to cover this and draw a map in a future post.)
While the desktop market has historically been driven by x86 architectures, the embedded market is much more fragmented. Some embedded licensees specialize on one particular architecture, like Intel, who only produces chips based on Intel-created IP. Others take an agnostic approach and license whichever architecture suits the solution they are seeking. Cavium Networks, for example, has produced products based on ARM, MIPS, SPARC, and x86 CPUs (and likely others I could not easily find).
All of the astute, intelligent, and frankly attractive readers of this blog remember that I am a Linux dude. I work in embedded systems, where there are tons of different architectures to choose from. Specifically, I work for a Linux provider who produces market-specific distributions and software development kits for a wide range of different architectures.
Linux has evolved to be a universal operating system. Originally conceived to run on x86 architectures, thanks to its open-source nature Linux has been ported to practically every modern microprocessor (and even an FPGA). It is particularly popular currently on ARM, MIPS, and PowerPC-based embedded systems, but for all intents and purposes, it can be made to run just about anywhere.
That sets the stage for a huge win-win scenario.
Wind River’s acquisition definitely fits, for lack of a better term, a “proprietary” model. As a standalone company, Wind River provided Linux solutions across the board just as MontaVista does. However, now they are owned by a company who is single-minded giant in desktop computing, and who probably acquired Wind River in the first place in order to extend their reach more deeply into the embedded market, where they have not had much recent success. There is no motivation for Intel to continue to provide software solutions for competitor’s products beyond current contracts, and a lot of motivation for them to close ranks. (Industry watchers will remember that this also happened to Metrowerks a short while after they were absorbed by Motorola/Freescale.)
However, the MontaVista/Cavium situation is different. Cavium is a solutions provider who is not tied to any particular platform. Cavium currently produces products primarily based on MIPS and ARM platforms (executives of both are quoted in the press release), but they are not limited to those, and have used others in the past, as stated above.
Those are the facts. What follows is conjecture based on those facts, simply my opinion.
This merger provides a strong impetus for Cavium to grow their product line into other markets. I would not be surprised to see Cavium extending its marketplace reach over the next few years, and expanding into new territory using MontaVista’s processor-agnostic philosophy as leverage. In fact, at the same time they announced this merger, Cavium also announced their intention to continue their relationships with Wind River and their other ecosystem partners, which I read as an desire to put a foothold on every aspect of embedded computing—ambitious, but now achievable. Cavium recognized a gap in their portfolio and filled it.
MontaVista, for their part, recognized a brass ring when offered, and took it. Other, more knowledgeable bloggers than I have already written about the merger in terms of MontaVista’s past, their business model, mistakes and successes. I am more concerned with the future.
The enormous potential is what I find so amazing about this merger. It provides MontaVista with a stable base where it can keep producing world-class software and tools, along with an extremely useful general embedded Linux community. It provides Cavium with an end-to-end hardware-to-software solution, as well as a set of software, tools, and ecosystem—support, QA, experienced FAE, professional services, documentation, community—to enter any embedded marketplace.
In other words, it is a win for both MontaVista and Cavium, and a huge potential upside for all of the customers of both companies.
UPDATE: Jerry Krasner at EMF agrees.