Xilinx vs. Altera: Which Is the Real Growth Stock?

In the semiconductor business, there are two niches that have had gross margins in excess of 60%. One of them is programmable logic. I worked for ten years in companies that wanted to emulate the success of Xilinx (XLNX) and Altera (ALTR). It took me ten years [and a $35 million dollar judgment against my last Programmable Logic employer] to figure out that this duopoly will own the programmable logic [FPGA and PLD] till the end of time [with a small market-share conceded to Lattice (LSCC), Actel (ACTL), QuickLogic (QUIK), Atmel (ATML) and Cypress Semi (CY)].

ALTR and XLNX make chips that can be programmed to perform any function you would want a chip to perform. Meaning, they make "blank slate" chips that can be configured to implement any logic function that you can dream of. In fact, Intel (INTC) uses huge boards made of tens of these chips to make sure that their logic works - before they get their CPUs fabricated into silicon. Smaller ASIC companies have become smarter over the last decade and make prototypes of their chips [using programmable logic] before they start mass-production.

The elegance of this solution is obvious, when you take into consideration the amount of money it takes to design a chip. By the time you design, commit to a mask-set, run a few wafers through TSMC/UMC/whatever, get the chips back and test them, you'd be out $10M. The only drawback to this approach is that the programmable chips run slower than a custom designed silicon solution, and one could not possibly mass-produce using the prototyping chips - as whatever gains you might have in time to market, are erased by the expense for these chips - that can cost up to $1000 a pop. If you use a PL chip, all you do is to erase its original configuration, load a new one, and repeat till you succeed [easy eh ??]. Yes, there are exceptions to every single statement that I made here [which is what makes analyzing semiconductors so interesting].

Companies that tried to compete with the two big boys include Intel, Motorola (MOT), AT&T (T), AMD (AMD). Needless to say, they either failed or divested their business. Actually, AMD deserves credit for creating the first PAL [through their MMI division - now part of LSCC]. The reason why all these brand-name companies tried to get into PL, is the gross margin number that I mentioned in the first sentence of this article. The reasons as to why they failed are manifold, but this article is about which of the companies is the better investment - looking forward to the next two to five years.

Both Xilinx and Altera recognize the fact that the business that they are in, however profitable it may be, is mature. Any attempt by either company to enlarge their SAM [served available market] always resulted in decreased gross margins - due to the fact that all programmable logic chips can be replaced by an ASIC. In fact, both ALTR and XLNX have a path that they offer - which will allow one to go from a "changeable" programmable logic chip to a "fixed" ASIC, but they both do it for a reason [Bapcha at Gmail if you want to know why]. They both generate prodigious amounts of cash, and pay a small dividend that they easily earn - as they recognize the fact that they are in a niche in semiconductors that is not expanding at the rate that it was in the last decade.

Looking at the prior quarter's numbers, Altera had revenues of $359M and gross margins of 67%. Xilinx's numbers were $488M and 64% respectively. Their revenue growth rates were 12.5% and 9.4%. ALTR bought back 15% of their shares vs. 7.26% for XLNX [yoy fully diluted]. EPS was 32c/share in 2008 and 22c/sh in 2007 for ALTR. XLNX's earnings were 30c/share in 2008 and 28c/sh in 2007 [last reported quarter]. Dividend was raised by ALTR from 4c/sh to 5c/sh yoy/qtr. XLNX went up from 12c/sh to 14c/sh.

Before looking at the numbers, I thought that my decision would be easy, but it sure wasn't. While Xilinx has always had an edge over Altera in terms of market share, profitability, and earnings predictability [historically], ALTR is growing faster, has slightly better gross margins, and most significantly, better employee morale. Xilinx might be a better stock for total returns as it pays a significantly higher dividend and has a CEO in place who is determined to contain costs, but it also has a work-force with a lot lower employee morale.

I started writing this article wanting to pick Xilinx over Altera, but I have to pick the real growth stock among the two - ALTR. The market seems to agree with me - rewarding ALTR with higher multiples than it does for Xilinx. When the semiconductor market heats up, both these stocks have traded at PS numbers in the double digits [but that was when they were real growth stories].

Bottom-line: ALTR for growth.

Disclosure: I have NO positions in any of the stocks mentioned. Statements about employee morale constitute my opinion. Both companies have employees that are happy and vice-versa. Both companies have significant software investments, but it is a means to an end [selling chips].

Comments

[daniela:]

Wow. Great analysis. Thank you.

2008 Aug 11 05:03 AM

[laguna_b:]

Employee morale? Both companies have the excitement internally of a grave yard. Neither company is able to take necessary risks to grow their parkets. They just continue to slice up the existing markets. Look at the results. Altera is trading in the same band today as it was 5 years ago and more. Both companies spend fortunes trying to keep up in process technology and nickel and dime real innovations internally.

2008 Aug 11 07:42 AM

[Alan Brochs...:]

How many angels can fit on the head of a pin?

I actually think both could do very well for a trade here. INTC looks set to go too. There are some real pigs in the sector that probably won't attract investor interest (SNDK, NVDA, MU), but several, including the FPGA guys and the HPAL guys look solid.

2008 Aug 11 09:08 AM

[btdt:]

Ironically enough, AMD owned 25% of Xilinx at one point and sold it off in the early 90's when they thought they'd realized a gain on their investment (XLNX has only split 4 times since then).

Enough of ASICs. ASIC design starts are a shadow of what they once were, and it hasn't been enough to propel the programmables into the stratosphere they were certain they were destined for. What DID happen was the rest of the logic market--ASSPs. Networking and other sectors got big enough that standard products emerged. Broadcom, for example, succeeded in a space that Xilinx thought it was going to own--but Xilinx had their eye on the wrong ball.

2008 Aug 11 11:24 AM

[PSLUNLIMITED:]

And the story goes that many years ago when the co-founder of Xilinx - BERNIE VONDERSCHMITT - worked for Zilog and proposed that they should seriously look at programmable logic he was turned down. The once venerable Zilog is now barely staying in business.... It is right up there with the Western Union of old turning down the purchase of AT+T as they saw no future in voice communication at the first part of the 20th Century! Great Article!

2008 Aug 11 01:37 PM

[FPGALover:]

Your investment decision should be made based on your time horizon. The challenge with the PLD business is that it takes 5 years from the time a decision is made to build a new chip and the time revenue ramps up. At least two years to introduce the new chip and another 3 years for customers to accept the new technology, build it into their complex systems (that thus require PLDs), qualify those systems, and sell them to their end customers. With this in mind, what we're seeing today is the impact on revenue of their 90-nm lines where Xilinx screwed up handsomely with Virtex-4. 65-nm was introduced in 2006 where Xilinx, for a full year, reaped up design wins with no real competition. It was Altera's time to screw up with Stratix III. Which means that sometime in 2009 the revenue trend should reverse. To reverse once more two years later when 40-nm technology ramps up into volume production, where Altera probably has an edge, thanks to TSMC... Nothing rocket science here.

But keep in mind that in the long term, the market cap of these companies tend to trade in a somewhat narrow channel...

2008 Aug 11 07:16 PM

[superal:]

Both companies have poor morale. To understand that better in the context of Altera visit whenlayoffsgowrong.blo....

2008 Aug 11 07:19 PM

[murf:]

Soon after joining the company this year he articulated how it now costs $70 million to design a new ASSP, using the latest process node. This requires a billion dollar plus market for such an endeavor to be viable.

Xilinx believes they need to focus less on being a technology innovator in order to create market opportunities, but instead pursue a business platform leadership position. This strategy appears to make sense since the last couple years analysts have pondered why ASIC designs have not seen a mass migration to FPGAs as expected. Instead ASIC starts have migrated more to ASSP products. This creates a huge opportunity for platform FPGAs to take market share from mid to low volume ASSPs.

2008 Aug 12 09:45 AM

[murf:]

seekingalpha.com/artic...

2008 Aug 12 09:46 AM