youtube
Xilinx CEO Victor Peng

During a December quarter in which many chip developers saw little or no annual revenue growth as an industry downturn arrived, Xilinx (XLNX - Get Report) grew its revenue 34%.

And though its growth is likely to slow a bit, analyst estimates, which might well be conservative, still call for Xilinx to see double-digit growth in fiscal 2020 (which ends March 2020) -- a time period when many peers are at least initially expected to see negative growth thanks to some mixture of inventory corrections, slowing end-market demand and trade/macro concerns among chip buyers.

Check out Eric Jhonsa's new tech blog, Tech Check, on Real Money, our premium site for active investors.

Why is Xilinx outperforming its peers to such an extent? It has a lot to do with an expanding addressable market for field-programmable gate arrays (FPGAs) and other programmable logic chips whose circuits can be reconfigured after the chip has already been manufactured, as well as a cyclical upswing for a market that has long been a major buyer of FPGAs.

Xilinx, together with Intel's (INTC - Get Report) Programmable Systems Group (the product of Intel's $16.7 billion 2015 acquisition of Altera), has long dominated the FPGA market and is also a major supplier of other types of programmable logic chips. Compared with chips whose circuits can't be reconfigured after manufacturing -- for example, CPUs and GPUs that can run a variety of applications, or application-specific integrated circuits (ASICs) created to handle a particular task such as decoding video, routing network traffic or running an AI software framework -- FPGAs have both strengths and weaknesses.

Relative to the ASICs and application-specific standard products (ASSPs) that they often compete against, FPGAs tend to feature lower processing power densities (that affects chip size, performance and power consumption) and maximum clock speeds. They also tend to be more expensive for use cases in which tens or hundreds of millions of chips are needed.

On the other hand, whereas it can take up to two years to develop an ASIC and tape out its design for volume production, developers can configure an off-the-shelf FPGA in a short amount of time to handle a particular task. And for low-volume use cases, an FPGA is often cheaper to use, since the R&D costs for creating an ASIC design and getting it ready for mass-production can be considerable.

For these reasons, FPGAs have been widely used in situations where developers need to quickly deploy chips optimized to run new algorithms or other code, and/or in which the code will need to be quickly updated one or more times after deployment. They're also why FPGAs and other programmable logic chips have been widely used in hardware testing equipment, and to prototype chip designs.

And in recent years, as developers of many stripes push the envelope regarding how fast cutting-edge algorithms and software platforms and standards are refreshed, the number of use cases for which FPGAs have become at least one of the popular options has been trending higher. Some notable examples include processors for driver-assistance systems (Xilinx says it's the No. 2 player here, behind Intel's Mobileye unit) and server accelerators that perform AI inference (the running of trained AI/deep learning models against real-world data and content).

There has also been a growing addressable market for system-on-chips (SoCs) that feature some mixture of programmable circuitry, CPU cores and hard-wired processing engines dedicated to specific tasks. In the December quarter, Xilinx saw revenue for its Zynq SoC line rise about 80% annually.

Later in 2019, the rollout of the first chips based on Xilinx's powerful ACAP platform (unveiled last year), which supports many different types of processing engines, will give it another way to grow its exposure to this space. Notably, the chips, which will be sold under a new product family called Versal, will leverage Taiwan Semiconductor's (TSM - Get Report) cutting-edge 7-nanometer (7nm) manufacturing process.

Meanwhile, the mobile infrastructure market, where programmable chips have long been frequently used to handle both radio subsystem (RF) and core signal-processing (baseband) applications, is seeing demand pick up as 5G rollouts commence. And Xilinx, which saw its sales to "Communications" clients rise 41% last quarter, has argued that 5G build-outs represent a bigger opportunity for it than 4G build-outs, thanks to increases in both the amount of hardware that will be deployed and the amount of chip content it can provide for infrastructure equipment.

Combine these secular trends with solid execution that appears to be allowing Xilinx to gain a bit of share from Intel's PSG, and the fact that the company doesn't have significant exposure to a soft smartphone market, and it's not hard to see why Xilinx is delivering healthy double-digit growth, even as many other big-name chip developers have seen, or are about to see, growth turn negative.

Certainly, a lot of these positives are now priced into Xilinx's stock, which has surged 23% in the two trading days following the release of a blowout December quarter report. Shares now trade for 29 times a fiscal 2020 EPS consensus of $3.80 -- a healthy premium relative to what many other chip developers now trade for. For that reason, investors bullish on Xilinx's growth story might want to wait for a better entry point.

At the same time, it's hard to deny that Xilinx's growth story is looking like a pretty unique one among chip developers at this point in time.

Check out Eric Jhonsa's new tech blog, Tech Check, on Real Money, our premium site for active investors.