I originally wanted to title this blog, “What defines a semiconductor company anymore?” It is no longer clear to me because more and more system companies, big and small, are taking the plunge, and becoming a member of the Do-It-Yourself (DIY) club. My email inbox certainly seems to suggest this trend is picking up.
Don't take my word for this though. Look at what is being said by the companies that play in this arena.
Last week I listened to TI's investor conference call, where they projected an 8 percent sequential drop in revenue for their next quarter. TI, the market leader in analog semiconductors, played down the projection, but what amazed me more was that their R&D investment as a percentage of sales is ~11 percent, whereas TI's competitors Maxim, Analog Devices, and Linear Technology are investing, respectively, ~22 percent, ~19 percent, and ~18 percent during these trying times. One of the analysts on the call questioned TI's logic behind lowered R&D investment at a time when sales are falling.
To get some more insight into this gloomy fourth-quarter outlook, I read TI's most recent annual report. Buried in that report they disclosed a developing industry trend whereby certain “…large customers are increasingly developing their own custom chips.” This made me wonder what is going on with Cadence.
Cadence sells what is arguably the premier set of analog IC design tools for the IC industry. Design tool sales give insight into the direction of R&D investment for integrated circuit design, since for every IC design engineer a company hires, they need another IC design software license.
In Cadence's quarterly conference call last week, Cadence's CEO Lip-Bu Tan expressed optimism for his company, claiming that he is seeing “…tremendous growing in the system companies going vertical.” Bear in mind, Cadence sells not only design tools but also analog IC intellectual property (IP).
Then we have Cirrus Logic, a high-performance analog semiconductor company with 82 percent of its 2013 sales designed into Apple products. Heck, Apple should just buy the company and either spin off or sell to TI the other 18 percent of Cirrus' business. In one year, Apple would probably be at break-even having saved the entire Cirrus markup.
This brings me back to the question, “What (or who) is a semiconductor company anymore?”
When I started designing analog ICs the analog companies manufactured their own wafers. They generally wrote their own design tools. They designed all of their circuit IP.
Today, just about any company selling an IC with an embedded processor is buying the IP from ARM Holdings. In short, any company with R&D funds can develop their own IC. Even those companies starting out in their garage! So what is left anymore that defines a semiconductor company?
Well, in a primarily analog IC company, at best only two things: product marketing managers who decide what products to design and the large pool of engineers who design those products.
So what's wrong with this picture?
For one, marketing managers spend their time trying to predict what their customers, the system companies, want to buy. Put another way, the very companies Cadence and TI claim are increasingly designing their own ICs — the vertically integrated system companies — know exactly what is required for future semiconductor devices.
It is often said that only 10 percent of the product portfolio on an analog semiconductor company's website actually turns into long-term profitable sales. The rest of the products are presumably failed predictions. Doesn't that mean that 90 percent of a semiconductor company's R&D is wasted on wrong predictions? This sounds like the venture capital business model.
And then we have the design engineering pool that is available to any company that pays competitively while offering a technically challenging and positive work environment. That compensation part is a lot less burdensome for a system company whose revenue is based upon the much higher valued, top-level system product. Apple, for example, spends only ~2 percent of its revenue on R&D.
While changes in an old industry do not happen overnight, I am seeing this design shift more and more. If the tool companies and independent IP providers had their way, they would help all system companies develop their own unique IC products. This is setting up a horserace between the traditional IC companies that now envision a future where they sell embedded processors, analog, and software; and the system companies that have traditionally purchased those same pieces, but from multiple providers.
This scenario reminds me of playing [American] flag football where one team, the IC companies, would be the skins and the other team, the system companies, would be the shirts (since shirts cover skins). Who do you think will win? And at what point should a system company decide to embark on building their own integrated IP?