中文 English

AMD Wants An FPGA Company, Too

Analysis: What Xilinx brings to the table, and what’s in it for both companies.

popularity

AMD signed a definitive agreement to acquire Xilinx for $35 billion in stock, setting the stage for a head-to-head battle against Intel in nearly all major markets. But there’s more to this acquisition than just keeping up with AMD’s arch-competitor.

To begin with, the acquisition has a big impact on the programmable logic market. The only pure-play FPGA vendors left are Lattice, Achronix, and QuickLogic, despite a growing need for programmable logic. That demand is reflected in the purchase prices. Intel bought Altera in 2015 for $16.7 billion. In contrast, Microsemi bought Actel back in 2010 for $430 million, and was in turn bought by Microchip in 2018 for about $8.35 billion. In addition, there are a handful of eFPGA vendors, including Flex Logix, Achronix, Menta, and QuickLogic.

Xilinx always was the one to beat in the FPGA world, and that gap could grow significantly if this acquisition wins approval. Xilinx’s programmable chips are used across a variety of market segments, including data centers where AMD is particularly strong.

Still, the apparent driver here isn’t about AMD wanting to get into the FPGA business. The real motivation appears to be a combination of platforms and programmable chiplets.

There are two main problems that programmable chips address.

First, it takes a couple years, on average, to develop a new processor architecture at leading-edge process nodes. During that time algorithms and software may have changed dozens of times. While it’s critical to keep pushing forward on those architectures, some flexibility has to be built in to stay current with software. Much of this can be done in software, but that has significant overhead in terms of performance, power, and security. Programmable logic that is tightly integrated with other processing elements allows the whole system to stay current with changes to existing algorithms, or entirely new algorithms. It’s still not as fast as an ASIC, but it’s much faster than developing a derivative chip at 7nm or 5nm.

Second, as new markets demand more customized solutions, having programmability options is a fast way to address those markets. AMD and Xilinx both have been working with advanced packaging approaches for the better part of a decade. And going forward, AMD needs Xilinx far more than Xilinx needs AMD. Its customers are demanding extremely high performance and a fair degree of customization, and that will only grow as edge computing begins to take shape over the next decade.

It helps that Xilinx is highly diversified. It was able to post strong results in fiscal Q1, despite expected slowdowns in consumer and automotive sales due to the ongoing pandemic. The company sells into mil/aero, industrial applications, automotive, broadcast, consumer, data centers, and wired and wireless markets, and in the fiscal year ended March 28, it posted revenues of $3.163 billion, up from $3.059 billion in fiscal 2019 and $2.467 billion in 2018.

On Xilinx’s side, the deal was simply too good to ignore. But as Shane Rau, analyst at IDC pointed out, the purchase price jumped $5 billion since the negotiations first came to light a month ago. During that time, AMD’s market cap has grown, making an increased asking price more do-able, and Xilinx has been able to negotiate a higher price for its stockholders.

“If you look at the numbers, Xilinx has been doing well in terms of market capitalization, but it has gone from about two-thirds of AMD’s valuation to about a quarter,” Rau said. “So capitalization has not grown as fast. From an FPGA market standpoint, both Xilinx and Intel PSG (Programmable Solutions Group, formerly Altera) are strong in high-end data centers. Companies like Lattice and Microchip play in other market segments.”

Stock market valuations are generating some enormous deals lately. The proposed $40 billion acquisition of Arm by Nvidia is a case in point. So is the pending $21 billion acquisition of Maxim Integrated by Analog Devices. As long as market caps remain high, those kinds of megadeals are likely to continue, with an emphasis on stock or mostly stock transactions. But when the stock market begins to soften, M&A activity likely will follow.

Related
Deals That Change The Chip Industry
Nvidia-Arm is just the beginning; more acquisitions are on the horizon.
Nvidia To Buy Arm For $40B
Analysis: What’s driving the deal, and what to watch for.
Good Vs. Bad Acquisitions
M&A begins to ramp up, but not all of them will work.
The Next Wave Of Consolidation
Economic conditions are all there. So what’s the holdup?
New Architectures, Much Faster Chips
Massive innovation to drive orders of magnitude improvements in performance.



Leave a Reply


(Note: This name will be displayed publicly)