AMD ranks as “major” chip player with Xilinx buy, including in the data center

Xilinx
Xilinx shares climbed 7% on news it will be purchased by AMD for $35 billion in stock. How the combined entity develops customized chips for servers will help dictate how well it can compete against Intel and a combined Nvidia/Arm. (Xilinx)

 

AMD may hope to clobber Intel in the data center chip market with its $35 billion stock purchase of Xilinx announced Tuesday.

That objective may take a very long while. However, AMD and Xilinx have great tech synergies in the data center (including with AMD’s fourth quarter production of Milan chips) and they both heavily endorse customization approaches, including using open source software.

Free Daily Newsletter

Interesting read? Subscribe to FierceElectronics!

The electronics industry remains in flux as constant innovation fuels market trends. FierceElectronics subscribers rely on our suite of newsletters as their must-read source for the latest news, developments and predictions impacting their world. Sign up today to get electronics news and updates delivered to your inbox and read on the go.

Such qualities will please server manufacturers that sell to large business customers and research groups wanting high performance computing.  Whether that puts the bigger AMD anywhere close in an actual race with Intel is debatable, but some manner of race is indeed underway.

“The combination of AMD’s CPUs, GPUs and now Xilinx FPGAs is a game changer for how AMD can compete with Intel,” said Jack Gold, an analyst at J. Gold Associates.  “It does up the game for AMD in being able to compete more broadly and it adds an installed base of Xilinx customers, many using Intel chips, that AMD can now attract to their base CPU/GPU designs.”

The acquisition effectively transforms AMD into a “major, mainline player in the broader computing space and not relegated to the low-cost secondary supplier it was in the past,” Gold said.  

Big picture, AMD and Xilinx combined would likely produce more than $11 billion in annual revenues across all chip categories, including chips for servers and PCs and laptops. That would rank the new AMD eighth behind Texas Instruments but still many times smaller than Intel, which dominates the industry and finished 2019 with $71 billion in overall revenues.

Whether it takes AMD a few years or many years to come close to Intel in revenues, if ever, a vital concern is how AMD with Xilinx can do in chips used in servers for data centers and high performance computing.

Intel had a down third quarter in the data center chip revenues especially with government and enterprises, while AMD doubled its third quarter revenues in that arena over the previous year.  

AMD’s enterprise, embedded and semi-custom segment revenue was $1.13 billion in the third quarter, an improvement of 116% year over year, including increased EPYC processor sales.

By comparison, Intel reported third quarter revenue of $5.91 billion in the Data Center Group, down by 7% over the previous year. In that group, Intel reported a 47% decline in revenues from enterprises and government after two quarters of growth above 30%.

“The data center environment is actually good,” AMD CEO Lisa Su said on a conference call early Tuesday.   “Data center continues to be very strategic and performs well for us.”

A big theme both she and Xilinx CEO Victor Peng underlined on the call is the combined companies’ ability to customize products for data center customers, which can include the use of open source software.

With server OEMs, Su said, “it’s very much a customer-specific message…There will be opportunities to do things on the hardware side and I view it as market customization. Doubling down on the software side is key to adoption.”

Peng added that initiatives in edge computing and edge cloud systems rely on a common software environment. 

“We share the view that the common software environment is so important,” Peng added. “Workloads want to customize things. We handle optimization and fast real-time for some heavy compute.” The addition of Xilinx to AMD means a “comprehensive powerful solution tailored to workloads and even endpoints and the cloud,” Peng said.

Su added, “We’ve both invested in the software environment and Xilinx has a strong software platform.” She didn’t identify any tools in particular, but Xilinx tags most of its development platforms  under the Vitis software brand.

Likewise, Peng said the Xilinx has already been collaborating with AMD on ROCm and other open development environment initiatives.

“Everyone relies on open source to succeed today,” analyst Gold said. “Open source is a key ingredient for everyone.”

Another analyst, Leonard Lee at neXt Curve, said the combined AMD and Xilinx will be more competitive against Intel but also against Nvidia with Arm.  Nvidia announced its intention to buy Arm from SoftBank for $40 billion in September and Nvidia said it plans to retain Arm’s open access model.

RELATED: Nvidia will buy Arm from SoftBank for $40 billion

Even as AMD will benefit in gaining a 5G capability with Xilinx as well as the Open RAN space with edge computing at the edge of the telco network, Lee said AMD is “relatively weak” versus Intel and Nvidia/Arm in the heterogenous compute and architecture race.

“The key for AMD will be to realize synergies early once the deal closes to pursue a broader platform/ecosystem strategy,” he said. Intel and Arm already well advanced in the market in such capabilities, Lee added.

“With Xilinx, AMD will have a great adaptive hardware story to tell” particularly with the addition of Xilinx FPGAs and hardware acceleration products, Lee added.

The announcement of the deal sent AMD stock down nearly 5% to $78.19 in early afternoon Tuesday trading, while Xilinx rose more than 7% to $122.70.

Suggested Articles

In June, Su stood out for heartfelt commentary on social injustice after the killing of George Floyd. She challenges industry to do more.

New chip will be incorporated in coffee mug-sized device to cost $199 next year

In a nutshell, Edge ML is a technique by which smart devices process data locally using machine and deep learning algorithms.