latest-news-headlines Market Intelligence /marketintelligence/en/news-insights/latest-news-headlines/analysts-question-technology-advantage-of-amd-s-xilinx-acquisition-60945362 content esgSubNav
Log in to other products

 /


Looking for more?

Contact Us
In This List

Analysts question technology advantage of AMD's Xilinx acquisition

Blog

Investment Banking Essentials: April Edition

PODCAST

Episode 10: Cloud Pricing & Complexity

Blog

OTT players find great value in partnerships but still prefer direct-to-consumer

Blog

Tracking Credit Risk of a Major U.S. Retailer


Analysts question technology advantage of AMD's Xilinx acquisition

Advanced Micro Devices Inc.'s acquisition of Xilinx Inc. could open new markets for the chipmaker, analysts said, but some questioned whether the technology AMD is acquiring will have as great an impact as its executives believe.

Intel Corp., for instance, has pursued a similar strategy with limited success in expanding its data-center market share, one noted.

AMD announced the $35 billion deal on Oct. 27. It said the all-stock transaction will increase the value of the combined company to $135 billion, increase AMD's addressable market to $110 billion and be immediately accretive to AMD's margins, cash flow and earnings per share.

AMD is interested in Xilinx's "adaptive computing" technologies, particularly its expertise in chip packaging and software designed to improve the performance of data centers by reducing the effort servers must spend on moving data and other administrative functions.

Those products, often referred to as smartNICs, can be "additive to our data-center portfolio today," AMD President and CEO Lisa Su told analysts during a call following the deal announcement.

SmartNICs and artificial intelligence accelerators are part of an effort to improve the performance of applications in data centers by adding components designed to add extra power to specific tasks that are slowing down a server or even a whole data center, according to Zeus Kerravala, founder and principal analyst at enterprise IT and network analysis firm ZK Research.

Xilinx is best known for making microchips called field-programmable gate arrays, or FPGAs, that can be added to data-center servers to improve the speed of AI applications, much like a graphics card can be installed in a personal computer to improve the performance of a game.

The ability to sell customized packages designed to accelerate individual applications or whole data centers can improve the ability of AMD and Xilinx to sell to both cloud service providers that require customized hardware and software for data centers as well as the enterprise data centers that need more standard technology to support their existing software and servers, Su said.

Using FPGAs to accelerate and adapt the functions available to data-center servers is an approach Intel has tried with limited success since its $16.7 billion acquisition of Altera Corp. in 2015, said Matt Bryson, senior vice president at Wedbush Securities.

AMD offers the kind of graphics chips that are the traditional way to accelerate data-center software, Bryson said. Xilinx gives AMD an option it can use to counter the advanced data-center performance products touted by Intel and NVIDIA Corp.

SNL Image

The addition could help AMD accelerate its growth even more, following a quarter during which AMD's data-center revenues rose by double-digit percentages. Intel, with consolidated revenue of more than 10x that of AMD, saw its Data Center Group revenues fall 7% during the most recent quarter and 47% among enterprise data-center customers, Bryson noted.

AMD's overall market capitalization has been rising along with its revenues during 2020, in part due to Intel's product development stumbles, which have left Intel trailing the technical advances of rivals including AMD. As of Oct. 26, the day before the Xilinx deal announcement, AMD had a market capitalization of $96.54 billion, still significantly less than Intel's $191.46 billion. Meanwhile, Nvidia this year surpassed Intel to become the largest semiconductor company by market capitalization, at $324.7 billion as of Oct. 26.

But Xilinx's current crop of FPGA-based accelerator cards lag behind the performance boost available from dedicated AI accelerators from other companies, which may be why they have generated tepid interest among data-center customers, said Linley Gwennap, president and principal analyst at The Linley Group.

"Xilinx's CEO sees the data-center push as a transformative opportunity for the company," Gwennap said. "But it hasn't made much of a dent in revenues and hasn't demonstrated a strong value proposition that would get its new products into the data center."

Xilinx does have an updated version of the products, branded Versal, that will come out in early 2021, with higher performance numbers and the ability to be more easily programmed to add financial forecasting, molecular research or other functions to existing servers. Those functions represent niche markets unlikely to generate mass sales even in specialized data-center markets, however.

"If [Xilinz CEO Victor] Peng were confident his Versal strategy would succeed, he wouldn't be willing to sell the company right before the revenue starts to flow," Gwennap wrote in a note immediately following the deal announcement.

"I assume the reason to buy Xilinx is to acquire a company that is already successful, with a broad revenue stream that might be able to help AMD in the data center," Gwennap said. "But it's not a great way for AMD to get into AI acceleration. Other companies have a better value proposition that AMD could have bought for a lot less money."