Many technology investors know about the rivalry between Intel (INTC) and AMD (AMD). Relatively few know or understand the rivalry between Xilinx (XLNX) and Altera (ALTR). The difference is that Intel and AMD are brands that are associated with PCs, for which there is consumer awareness. Xilinx and Altera make PLDs (programmable logic devices), which are incorporated into many non-consumer (and some consumer) electronic devices. You don't see their brands highlighted in stores.
Yet as the PC market has shrunk the PLD market has held up and even grown in some market segments. PLD growth seems likely to continue. Here I'll focus on Xilinx's Q4 2013 results and its likely performance in 2014, with implications for the direction of the stock price.
While there are other players in the PLD market, Xilinx and Altera revenues can act as a proxy. In Q4 2012 their combined revenues were $949 million. In Q4 2013 combined revenue was $1.041 billion, an increase of 9.7%. Contrast the combined revenue of AMD and Intel, which was $14.7 billion in Q4 2012, but $15.4 billion in Q4 2013, up just 4.5%. Intel did worse, with Q4 revenue up just 3% from Q4 2012.
Xilinx had Q4 revenue growth (from year-earlier) beating both the (mainly) x86 / PC graphics trends and the PLD trend rate. Revenue was $587 million, up 15% from $510 million in Q4 2012. However, it was seasonally down 2% from $598.9 million in Q3.
On a GAAP basis, Xilinx net income was $175.9 million, up 24% sequentially from $141.5 million, and up 70% from $103.6 million in the year-earlier quarter. Diluted EPS (earnings per share) was $0.61, up 24% sequentially from $0.49, and up 60% from $0.38 year-earlier.
The stock price has reflected, at least in part, the success. Xilinx had a 52-week high on 1/22/2014 of $48.91. The 52-week low was $34.98 on 4/19/2013.
There are a number of specific trends underway that could to be examined to determine if Q4 is a fluke that will revert to some lesser mean, or if it is an indicator of continued strong revenue and profit growth. For Intel, most revenue comes from x86 processors, and subsectors of the market are servers, desktops, notebooks, tablets and smartphones. In contrast Xilinx markets cover 3 broad industrial segments: Communications and Data Center; Industrial, Aerospace & Defense; Broadcast, Consumer, and Automotive; and other.
Xilinx has a number of product lines, but in general they cut across the end-customer segments: Virtex, Kintex, Artix, and Spartan are all FPGAs (Full Programmable Gate Arrays). Zynq is an SoC (System on Chip) that includes both an FPGA and dual ARM microprocessors. Using Kintex as an example, some of its applications are in avionics, displays including 3D TVs, LTE (4G wireless) baseband stations, and various video applications including high-end cameras. The ability of PLDs to be programmed at the hardware level gives them an advantage wherever there is a need for speed combined with a limited set of repetitive calculations to be made, as when there is an audio or video data stream being processed.
Breaking out end-use segments by dollars for Q3 and Q4 2013 we can see the short-term trends against the seasonally slower Q4:
|segment||Q3 2013, $ million||Q4 2013, $ million||change|
|Communications and Data Center||257.5||258.2||0.2%|
|Industrial, Aerospace & Defense||227.6||217.1||-4.6%|
|Broadcast, Consumer and Automotive||95.8||93.9||-2.0|
Two sectors differed substantially from the seasonal sequential drop. Industrial, Aerospace and Defense saw a drop of 4.6% that may have more to do with the sequester and government shutdown than with long-term demand. It is a segment that should be monitored, since it accounted for 37% of Xilinx revenue in Q4.
Communications and Data Center is the largest segment, accounting for 44% of revenue in Q4, and did well against a background of slightly negative seasonality. The key variable within the segment at this time is PLDs for 4G base station rollout. PLDs can be used in smartphones, but those tend to be low-end PLDs because they only need to handle single-device bandwidths. Base stations need to be able to handle signals to and from numerous phones, so the newer models tend to be designed in when possible.
Currently, while 4G is still being rolled out in the U.S. and Europe, the big rollout is in China, where it began in earnest in Q4. The main source of incremental wireless sales within the Communications segment was 4G in China. However, segment sales were broad-based, with continuing sales into other nations and 3G. The Chinese rollout of 4G base stations is going to take several years; it should not just be a Q4 spike. 4G rollout is mostly in the future in developing nations, including India, so the demand picture looks quite good for the foreseeable future.
Another way to dissect Xilinx revenue is by the age of the product families. Base products are the oldest, and were typically designed into products 2 or more years ago, though they could be used when low-cost components are needed in new designs. They represented 26% of revenue in Q4. Mainstream products have been around for a while but are no longer cutting-edge; they represented 33% of revenue. New products like the Zynq 7000 are still being added to many new device designs in, but they have ramped to 38% of revenue. The remaining 3% of revenue comes from software, support and services.
Process nodes are as important to the PLD industry as they are to the rest of the semiconductor industry. Xilinx's newest products are at 28 nm, but 20 nm products are now being sampled. Generally lower nanometers means more processing capability with lower power requirements per instruction executed. Xilinx reported that 28 nm products are ramping faster than expected and hit over $80 million in Q4.
In general the trend appears to favor PLD revenue growing faster than overall semiconductor revenue, though not as fast as ARM chip revenue has grown since the introduction of smartphones. Usually the main concern about Xilinx is that it might lose market share to Altera. In 2013, however, Xilinx revenue grew faster than Altera's (15% growth vs. 3% growth, Q4 to Q4).
Altera's argument for future market share gains lies mainly in its plans to go to Intel's 14 nm process for the next node, while Xilinx will use TSMC's (TSM) 16 nm node. Intel's node is likely to be ready earlier and in theory can pack in more transistors on any given die size. However, there are a lot of variables that determine the choices end customers make. Since both companies still get most of their income from nodes above 28 nm, and are just introducing 20nm products, the 14 nm vs. 16 nm is not likely to have much impact on revenue until 2016. Until engineers can see what these future designs can do, the most probable assumption is that PLD industry-wide sales trends will be more important than any small market-share jockeying between Xilinx and Vertex.
I don't see any highly probable reason Xilinx revenue would not continue to grow in a 10% to 15% annual range in 2014. Neither does Xilinx management: guidance for the March quarter is sales between $598 and $622 million, which would be up 2% to 6% sequentially. Macroeconomics could affect demand, but while some analysts see China as struggling, I would be surprised if the 4G rollout gets held up.
Since we are in a market dip it is a particularly good time to eye Xilinx for addition to portfolios. At the February 3 closing price of $45.02 Xilinx has a market cap of $12.0 billion at, and a dividend yield (at $1 per year, $0.25 per quarter) is 2.15%. Xilinx went ex-dividend on February 4. It has a trailing P/E ratio of 21.1, which is reasonable given growth prospects in 2014.
However, Intel currently has a P/E ratio of 12.7 and a dividend yield of 3.7% (at $23.95). So to an extent the gloom about Intel and hopes for Xilinx have been priced into the market.