Demand for artificial intelligence (AI) chips is going to go through the roof in the coming years. Allied Market Research estimates that AI chip demand could increase at almost 50% a year through the next five years, as this technology gains adoption across several verticals ranging from automotive to healthcare.
Xilinx(NASDAQ: XLNX) is well aware of this massive opportunity, which is why AI-focused development was a common theme at its recent analyst day. The company has already taken its first steps in the AI space with the help of its field-programmable gate arrays (FPGAs), but it's now looking to raise its game by going all in on product development.
Man holding a tablet that's projecting an array of lights shaped like a human brain.
Image source: Getty Images.
Xilinx will pour more money into R&D
Xilinx has been stepping up its research and development (R&D) efforts over the past few years. It spent around a quarter of its revenue on R&D last year, up from 20% four years ago. In fact, the chipmaker's R&D expenses have been growing at a faster pace than its top line for three years now.
This trend is expected to continue in the new fiscal year, as Xilinx plans to increase its operating expenses nearly 10% in fiscal year 2019, which will be identical to its revenue growth. Quite clearly, Xilinx isn't focusing on boosting profitability right now; it's looking to accelerate its product roadmap to deliver more efficient chips.
The majority of the company's chips are currently based on the 28-nanometer (nm) and 20nm platforms, but it is planning to move to the more efficient 16nm and 7nm nodes. Chips manufactured using smaller nodes are more efficient because they can deliver more performance for each watt of power consumed. Additionally, they tend to have lower manufacturing costs, since the chipmaker can manufacture more integrated circuits from each wafer.
As such, Xilinx's manufacturing costs should come down once it starts mass production of such chips, leading to greater profitability. But this is just one side of the coin; making FPGAs using a more advanced manufacturing process should help Xilinx tap the lucrative AI market.
Where's the money going?
Xilinx has set its sights on the data center space, which is going to witness a massive spurt in demand for AI chips, as the cloud is going to handle the majority of AI workloads. In fact, Xilinx estimates that demand for data center chips will increase at a compound annual growth rate of 67% over the next five years, hitting $4.6 billion in value, which is nearly double Xilinx's annual revenue.
Chips needed to compute AI workloads will account for the majority of this end market, which is why Xilinx is busy rolling out its 7nm-based Everest chips. Xilinx claims that these chips can compute AI workloads at least 20 times faster than its current 16nm chips. The company has started testing the new chips with select customers, and plans to start shipping them in 2019.
This sets Xilinx on its way to gain ground on archrival Intel, which has delayed its jump to the 10nm manufacturing process. Chipzilla will remain stuck with the current 14nm process for the foreseeable future, so Xilinx should be able to consolidate its FPGA lead. As it turns out, Xilinx already enjoys a strong technology lead over Intel, which is why it controls close to 60% of the FPGA market.
Allied Market Research estimates that FPGAs will be the fastest-growing sub-segment in AI chips, outpacing even graphics cards from the likes of NVIDIA. This is because the flexible architecture of FPGAs means they can be reprogrammed by developers for other tasks.
Moreover, compared to power-hungry graphics cards, FPGAs are power-efficient. This means they're more suitable for deployment in large-scale scenarios, such as servers, where keeping energy consumption low is key. Not surprisingly, several cloud service providers have started deploying FPGAs in the cloud.
Amazon, for instance, partnered with Xilinx late last year to deliver an FPGA-based cloud computing instance. But this isn't the only big cloud partnership that the chipmaker boasts. Baidu and Alibaba have also selected Xilinx FPGAs to accelerate their cloud services.
A smart bet
Xilinx's revenue growth hasn't been very attractive over the past few years, but that was before the AI use cases for FPGAs emerged. Cloud service providers are using these chips now, and with promised technological advancements from Xilinx, it won't be surprising if they gain broader adoption to facilitate AI deployments.
Once that happens, Xilinx's expenses should come down as a percentage of total revenue, boosting its earnings. This is probably why analysts expect the company's earnings to grow at a CAGR of nearly 12% over the next five years, well ahead of the 4% annual bottom-line growth it has clocked over the past half-decade.
The opportunity it faces and a forward P/E of 24 compared to the industry average of 30 make this stock a smart bet.
John Mackey, CEO of Whole Foods Market, an Amazon subsidiary, is a member of The Motley Fool's board of directors. Harsh Chauhan has no position in any of the stocks mentioned. The Motley Fool owns shares of and recommends AMZN, BIDU, and NVDA. The Motley Fool has a disclosure policy.
