Eventually short-term returns are not possible without sustained R&D.
Eventually the second tier reaches 28 nm. If no one demands post-28 nm technology, because it's too expensive, certainly the habit of lowering costs will lead to the end of semiconductor technology leadership.
@resistion: I agree... in the longer run this trend will kill the Golden Goose that is giving the short term gains in the first place! The Semi industry is in a quandary -no one seems to have a clear strategy that nourishes a healthy R&D in the long term while keeping the short term-minded shareholders happy! And I don't see how the companies can do this while the industry is deeply entrenched in a suicidal pricing model.
@Rick: seems to me that the next generation technologies like Compound Semi, Graphene(?), etc., will have a difficult time attracting investors who are in for the long term. Many of these newer technologies are where Silicon used to be three decades ago. I do know that the US Government has been funding some of these but there needs to be investments with a long term vision for these from the commercial sector.
@Doc: I suspect there will be at least some federal R&D dollars for the big next leaps (graphene, nanotubes, etc) though researchers would probably say it is not enough or as much as govts outside the US spend.
@Rick: I met with an engineer from National Renewable Energy Laboratory (NREL, www.nrel.gov) recently at SemiTherm 2014 who was mentioning about increased funding for compound semi's from the US Government. You are right, it is not enough. It is tough to sell the value proposition to private investors on these new technologies which enter at much higher cost than their Silicon-equivalents.
It's perfect timing for our rivals to catch up upon us. But so be it. The US should always focus on the top profit margin stuffs. As for the semi, whose profit margins are dropping year by year, let the Asians do it.
Interesting thought @lidation but I don't agree, silicon is strategic...I am sure Asia wil happily take over semi business and start selling us chips in increasing quantities...and we should move on to what? social networking? financial services? artificial intelligence?
@krisi, I was being a bit sarcastic. I am truly aware of the strategic aspect of the silicon industry to our national security. However, sadly, what is happening is happening. I am with one of the major semiconductor manufacturing equipment companies. Most of our business are in Asia since nearly a decade ago because the manufacturing happens there.
We like it or not, the Wall Street folks don't care. Money talks - unless the government steps in.
I see your point @lidation...all manufacturing even for highly complex silicon chips is moving to Asia...and apparently many people in US are fine with this, I read this article in Bloomberg that this is a natural evolution like agriculture going away many years back...but I just have difficulty imaging this service society where you don't produce anything...your suppliers can squizze you anytime they want in such a situation but not seliing you chips or increasing prices 10 times...what are we going to do then? Kris
This behavior is typical when Management considers the business to be in the "cash cow" stage, meaning that it is a mature company with chances for innovation and differentiation largely behind it. This fits with an industry where generational improvements are increasingly expensive and the return on them are smaller.
From a business point of view the logic of this is defendable. The chip industry has defined innovation as smaller geometries and faster operation. Lately this has been modified to include lower power, but it is still a pretty narrow range of innovation. When is the last time that there was a fundamental breakthrough in semiconductors? What are the chances of that happening in the future?
Yep, semis are now cash cows (if you are luckly enought to be generating cash). The big issue, I think is that semiconductors are now longer creating massive new markets to sell into. Will the Internet of Things be the next mass market? Not really since the semi content for the end nodes will be very small. Maybe the big data generated will create some additional server and networking infrastructure, but not the equivalent of the massive new markets (PCs, Networking, cell phones, etc) we saw during the good old days where we were growing in double digits....
There are still large potential growth segments:virtual reality which supposes to give amazing experiences and needs complex IO and high performance is one. Robotics(including household robots) is another possibility. And AI also seems like one.
But still , growing 10% from $10 billion is much easier than 10% from 300 billion.
And yes,IOT won't offer huge growth for the industry, but can offer big growth for mcu companies.
Probably its not all the companies are reducing the amount of investment in R&D. As may companies clearly aware it is not possible to reap the benefits without the present investment in semiconductor industry.
@Scott Elder: From what I see, TI has not reduced R&D so mich as maintained it at a current level while it increases the amount of its profits its returns to shareholders. It says it is gaining share in its corre target market of analog with a flat spend on R&D y-o-y.
Cadence's CEO will tell you that their company is seeing great sales growth from system companies whereas the semiconductor companies are flat to down. Who would know this better than a company that sells R&D tools to the semiconductor companies?
Touchstone's demise is really a story about the "old" semiconductor industry whereas Cadence is living with the new industry. Namely, for old school semiconductor companies, its back to selling $70,000 purchase orders to 100,000 people rather than $70 Million to one (Nokia, RIM, Apple, Samsung, etc.). It is a slow growth business model chasing $70,000 purchase orders one at a time.
Touchstone's business model was "the big guys are leaving the small guys". The problem with that strategy is that the customer's of the big guys have, in the interim, decided to DIY! (i.e. Google, Cisco, Microsoft, Apple, etc.) Only the small guys -- and their are 100,000 of them -- are left. So TI and others go back home to the small guys and Touchstone was back to competing with the old school.
Here is my prediction: The old school semiconductor companies will become the "Kickstarter" for new companies. Buy your complex silicon pieces cheap and quick. When you become successful, call Cadence, ARM, and TSMC and then go DIY.
If you been following the semiconductor industry for any time, you are aware that capital investment waxes and wanes. In a down economy for semiconductor electronics, we see companies cutting CapEx to bring expenses in line with lower revenues, and CapEx increasing again when the economy has an upturn.
The added issue we see now is a consequence of the predicted impending demise of Moore's Law. As process geometries shrink, costs rise. The amount of CapEx required to fund R&D and develop new processes is arguably rising exponentially.
We were seeing the precursors to this as increasing numbers of companies went "fabless", and others entered into joint ventures, because the cost of building fabs became greater than all but a handful of companies could afford. Similar concerns apply to R&D. Companies are still willing to invest in R&D, but how much they invest is hitting limits.
R&D investments in new technologies are gambles. The company making them is making a bet that the gamble will pay off, and they will develop new processes and techniques that will let them better serve existing markets and open whole new one. But the larger the bet is, the riskier it is, and you reach a point where a big enough bet might sink the company if it doesn't pan out.
No surprise companies are getting more cautious about the bets they place and the amount of R&D they will fund. Stockholder pressure for returns is only part of the reason.
It's ultimately about money. People make investments because they want to make money. They can do so in capital gains - the value of their stock increases - or in income from dividends paid on the stock, or both. But one way or the other they expect returns. The companies they invest in have similar realities to face. They want to survive and survival requires generating adequate returns on their investments, with much higher scrutiny of what investment is required, what potential returns are, and just how risky the bet is.
As traditional fabless companies cease their operations or reduce their R&D, their former chip designers are finding jobs at the fabless companies' former customers. Check out LinkedIn to see how many of your SoC architect/designer friends are working at Google, Apple, Microsoft, Facebook, Ebay, etc. instead of fabless semis.
There's economic logic behind this shift, and it isn't just a pendulum swing between verticalization and horizontalization. This presentation I gave at Semico's IP Impact conference makes that case (this links to PDF of my preso at SemiWiki): http://bit.ly/1guqEvi
@KurtShuler: thank you for the slides... if system houses like Apple can successfully incorporate (vertically) chip designs, why isn't the opposite model (of Chip design companies incorporating systems) not succeeding so far? Or, is it too soon to tell?
Fabless semi companies have traditionally had trouble creating "complete products" and selling them to a "market". I think it's because many of the fabless semis started out as custom ASIC houses, designing a chip based on a specification written by a single customer.
So, the DNA of a service provider or OEM is different than the DNA of a traditional fabless semi. The conundrum is that successful fabless semis will have to focus on creating "complete products" (including software, integration services, etc.) for narrow target market segments where the semi has established unique knowledge. A fabless semi that concentrates on no market segment and participates in all markets no longer has a competitive advantage. Anyone can design a chip with today's technology, tools, and value chain.
A Book For All Reasons Bernard Cole1 Comment Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...