Deep learning is a subsegment of artificial intelligence (AI) in which multiple computers to talk to each other through artificial neural networks. These artificial neural networks allow computers to understand random information without any prior programming, and Xilinx's (NASDAQ:XLNX) field-programmable gate arrays (FPGAs) are ideal for powering the deep learning of applications such as self-driving cars. FPGAs are integrated circuits that are programmed after they are made.

Xilinx's business has taken off remarkably since last year. The chipmaker has benefited from its technology lead over rival Intel (NASDAQ:INTC).

XLNX Revenue (TTM) Chart

XLNX Revenue (TTM) data by YCharts

More importantly, Xilinx isn't going to run out of steam anytime soon. The company is pulling the right strings in the right areas to attack the growing use of FPGAs, especially in the field of deep learning.

Man holding a tablet projecting an AI brain.

Image Source: Getty Images

How deep learning will drive FPGA growth

It is commonly thought that graphics processing units (GPUs) are necessary for training deep learning models as they can perform thousands of operations concurrently. CPUs are also in the mix, but they have a lot fewer cores than a GPU.

But it is believed that FPGAs could match or beat GPUs in deep learning thanks to their programmable nature. A FPGA can be reprogrammed to perform a variety of tasks after manufacturing, giving developers the flexibility to optimize them for specific models, which isn't possible with a GPU.

Furthermore, FPGAs deliver stronger  performance for each watt of power consumed. So they could be the chipset of choice in large-scale deployments such as data centers, where energy efficiency is key. As a result, FPGA deployment for deep learning applications is expected to gain momentum over the next few years.

Tractica's estimates suggest that FPGAs were almost non-existent in deep learning applications until last year, but they will equal (if not exceed) CPU deployments in this space by 2025. As a result, FPGAs could carve a greater share of the $12.2 billion deep learning chipset market by 2025, setting the stage for Xilinx to significantly boost its revenue given its lead in this space.

Xilinx is pushing the envelope to win the FPGA battle

Xilinx's biggest competitor in the FPGA space was Altera, which is now a part of Intel. But Intel has been behind Xilinx on the FPGA technology curve, and the latter is currently working to further extend its lead by moving to a smaller process node. Chips manufactured on a smaller process node ideally perform faster, consume less power, and are less expensive to manufacture.

This is why Xilinx claims to have built an 18-month technology lead over Intel by shipping more than 14 types of 16nm chips to over 450 customers. Xilinx had 200 customers  for its 16nm chips at the end of 2016, when Intel had just about launched its 14nm chips.

Therefore, Xilinx's chips are still in strong demand despite Intel launching a chip that's based on a smaller node. In fact, Xilinx's 16nm chips were on the market for a year already when Intel had commercially launched its own solution.

This technology lead has helped Xilinx increase its market share in programmable logic devices to 59% at the end of the fourth quarter of its fiscal year 2017. By comparison, the company had 52% of the market under its control at the end of 2011.

Furthermore, Xilinx can grab a greater share of this market as it has already started testing a chip based on the 7nm process. While Xilinx was announcing its 7nm chip in September this year, Intel disclosed that it has started the production run of test chips based on the 10nm platform. This indicates that Xilinx will continue to enjoy a technology lead over Intel in FPGAs, which is why it is confident of boosting its market share to a range of 60%-65% by  2021.

More specifically, Xilinx believes that the market share gains will add $250 million to its annual revenue by fiscal year 2021. Meanwhile, the secular growth in the end market thanks to the growing deployment of FPGAs in deep learning applications such as advanced driver assistance systems (ADAS) could add another $500 million in revenue.

By comparison, Xilinx has generated almost $2.4 billion in revenue over the trailing 12 months and top-line growth will eventually filter down to the bottom line. Analysts expect Xilinx's annual earnings growth rate to more than double in the next five years as compared to the last five, which indicates stronger upside potential going forward.

This article represents the opinion of the writer, who may disagree with the “official” recommendation position of a Motley Fool premium advisory service. We’re motley! Questioning an investing thesis -- even one of our own -- helps us all think critically about investing and make decisions that help us become smarter, happier, and richer.