Mega Trend 2: Escalating Software Content and Complexity
As smart connected devices penetrate every aspect of our lives, software content and complexity increase inexorably. The unabated increase in the amount of software in everything we use surely is one of the mega-trends of the technological revolution. Peripheral devices like printers or digital cameras have transformed into computers running sophisticated operating systems. The electronics content of automobiles now represents about 15% of costs, expected to grow to 25%, and those smart devices require an ever-increasing amount of software. The same thing is happening in military assets, transportation in general, on the factory floor, the office and now even the home. Smart connected devices of every shape and form are getting smarter by the minute, and “smart” is nearly synonymous with “software”.
I doubt many people would argue about the growing importance of software. The point of contention is who gets to benefit from the buildup of software IP, and who just suffers increased costs of production? In past presentations, WIND relied on the increasing complexity argument to justify its thrust into vertical markets. With so-called verticals, barriers to entry are erected with consequent increases in both market share and ASP. Revenue opportunities beyond organic sector growth were identified as coming from WIND’s primary, and only remaining significant, competitor: in-house developers. Multiple surveys seem always to confirm that WIND’s market share is slightly less than the total of what in-house developers represent. Most of us concluded that theoretically WIND could double in size before it would be forced to grow in synch with the embedded sector.
We now know that that analysis overlooked an important aspect of escalating software content. Everyone knows that a processor without software support isn’t worth the sand it is made of. This tautology is as true today as it was when the microprocessor was first introduced to the market. That’s why semiconductor companies have from the beginning thrown in free compliers, operating systems, and lots of other software to help their customers get their chips to market. Wary OEMs have always resisted these giveaways for fear of a hardware lock-in. Most OEM’s strongly prefer software support to be provided from an independent software vendor, especially one that enables the OEM to shift in the future to whatever hardware platform is most cost-effective. Moreover, attempts to combine software and hardware, forming a total proprietary solution started failing in the 1980s and became politically incorrect in the 1990s. The computer industry operates today mainly on the notion that no company can command a growing market position by combining proprietary software on top of their own hardware. There are exceptions in niche markets, like IBM’s mainframe computers and Apple computers. The other place exceptions occur is in non-commoditized, or immature, markets, ranging from the likes of Palm organizers to Cisco routers. One of the classic images that fell out of the Microsoft monopoly trial was that of Gates telling Andy Grove, “You do hardware, we do software, so stop developing NSP. [to paraphrase]”
Everyone understands each type of company’s position in an ordered Universe, so why are we feeling tectonic shifts in this 1990s-established business belief? Why are semiconductor companies suddenly investing greatly in proprietary software? Have the rules changed? If so, what does it mean for the embedded software juggernaut, WIND?
Look at Intel’s Internet eXchange Architecture as one example. The chip was announced in August of 1999, yet frankly there wasn’t much that could be done with it sans a library of callable microcode for the six onboard packet-processing engines. Interprocess communication, timing and any number of other subtleties slowed the chip’s deployment. While Intel ended up adding capability to the packet engines (no doubt because of software demands), most of the heavy lifting was, and probably still is, in getting all the low level software ready for serious production work.
The story must be the same with both IBM’s and Motorola’s network processor undertakings. In fact, the story is pretty much the same with most exotic multi-core chips. Software is becoming an increasing part of the effort to develop competitive chips. Software is becoming so important to semiconductor companies that they are crossing the point where they only have an interest in selling hardware. First, they want to benefit directly from such depths of development. In their minds, software is no longer a throw away, if only because they are spending a significantly share of their budgets on software. Second, by plunging so deeply into software, they actually are beginning to identify themselves as much with software as with hardware.
At issue, however, is not what semiconductor companies desire, or how much they think they deserve to be repaid for heroic software efforts. At issue, is reality. Any semiconductor company that wants to be a major player in network processors, exotic multi-core chips, reconfigurable chips, or anything else that demands an abundance of low-level software to be usable, must make their offerings compatible with the Tornado development environment and add-on toolset. No semiconductor company can afford to bypass the Tornado-based market place. With exotics, like multi-core, intelligent network processors and the like, this means either a one-time port, or a continuous, joint-funded development effort called a Center of Excellence, CoE.
With a new exotic announced weekly, one-time ports are relics of the past. This leaves CoE as the only practical alternative for both the semiconductor company and for WIND. WIND’s CoEs currently cover 75% of the embedded chip universe relevant to VxWorks, and new CoEs will be signed to extend the market coverage to about 90%. Anything more becomes cost-ineffective from WIND’s point of view.
Now put the pieces together. We know the software content of everything is escalating, including low-level software needed to make elementary sense of exotic chips. We know that all major semiconductor companies must, and most already do, share a joint development roadmap with WIND through a CoE, and that all successful exotics must be Tornado compliant. We know that semiconductor companies want to benefit from their growing software effort. Finally, we know WIND is a software company that now recognizes so-called “in-house” development is not limited to their OEM customers. (In fact, rough calculations indicate the amount being spent on software by semiconductor companies exceeds WIND’s previous estimates of the total embedded software market. Semiconductor software development represents a new and sizeable opportunity for WIND, made especially attractive by the ease of accessing the small number of semiconductor companies that count.)
Software content is escalating at every level in the supply chain, but WIND has an impenetrable position just above and independent of hardware that reaches up to and often includes middleware. Through its strategic relationships with OEMs like Sony, WIND owns the important customer relationship. Since most supply chains are being routed through the WIND tunnel, WIND is optimally positioned to be the primary beneficiary of the mega-trend of software escalation.
A metric that confirms this conclusion is embodied in the CoEs themselves, and how the relationships evolve with powerful semiconductor companies like Intel, Texas Instruments, Motorola, IBM and Hitachi. Look for most VAR deals and reference solutions to be negotiated mostly with semiconductor companies going forward, rather than middleware suppliers like Liberate, Flashpoint and Adobe. As CoEs and VAR deals with semiconductor companies firmly establish Tornado as the dominant platform, look for increasing signs of strategic relationships with major OEMs like Sony, Philips, Alcatel, Nortel, Cisco, Compaq, and now also companies like Intel and 3Com.
Pressures exerted by the mega-trend in software content show signs already of a steam boiler about to blow its top. It is hard to predict how each company participating in the silicon/software food chain will react to inevitable changes in market position and pricing power. Since semiconductor companies already are delving deeply into software, it is not clear how much farther they may be willing to go. We see them as hardware companies, but that it not necessarily how they now see themselves. Consequently, it is possible that any one of them could make a bold move, aided considerably by the weak economy and the change in the accounting of amortization. (For example Intel has $13 billion in cash.) WIND could end up being both victim and prize when the boiler blows.
Allen |
