AMD’s acquisition of Xilinx, the semiconductor industry’s ultimate chaos coming?

Another deal that will affect the landscape of the semiconductor industry.

According to Reuters, on October 27, 2020 AMD agreed to acquire Xilinx (Xilinx) in a stock deal valued at $35 billion. The purpose of the acquisition is clear – AMD wants to strengthen its fast-growing data center business.

AMD expects the deal to close by the end of 2021, with the two companies combining to have 13,000 engineers. The combined company has no manufacturing facilities of its own and relies entirely on TSMC to outsource production.

AMD Chairman and CEO Lisa Su will serve as CEO of the new company, and Xilinx CEO Victor Peng will serve as president, overseeing Xilinx’s business and strategic growth initiatives.

Under the agreement, Xilinx shareholders will receive each share of Xilinx common stock in exchange for approximately 1.7 shares of AMD common stock. Each share of Xilinx is valued at $143, approximately 24.8 percent above the closing price of $114.55 on October 26. As a result of the merger, AMD shareholders will own approximately 74 percent of the new company’s shares, while Xilinx shareholders will own approximately 26 percent.

On the same day, AMD also announced its 2020 Q3 financial results, and the numbers were stellar. On a GAAP basis, revenue was $2.8 billion, up 56 percent year-over-year; net income was $390 million, up 225 percent; earnings per share was $0.32, up 191 percent; and gross margin was 44 percent, up 1 percentage point year-over-year.

Strong demand for PC, gaming and data center products contributed to record quarterly revenue,” said Sutzer Fung. She is optimistic about the fourth quarter and expects year-over-year revenue growth of more than 25 percent.

For the fast-growing data center business

A spate of mergers and acquisitions has put the semiconductor industry landscape on the eve of major changes.

On September 14, NVIDIA announced it was acquiring Arm from SoftBank for $40 billion to shore up its shortcomings in CPUs, and in October, Intel divested its NAND storage business and sold it to SK Hynix for $9 billion to focus on its data center and PC businesses. And AMD’s acquisition of Xilinx is also an attempt to capitalize on growth and put more resources into data centers, where there is more room for growth.

Currently, AMD’s revenue pillar is the computing and graphics division. In the latest Q3 2020 results, for example, this division accounted for 59.6% of total revenue ($2.8 billion), or $1.67 billion.

Broken down, this division consists primarily of AMD’s Ryzen (Raptor) CPUs and Radeon GPUs, with the product vehicles being desktops, laptops, and workstations, or what we call the PC market. In the CPU space, AMD has been very strong lately, having just released the gaming-powered Zen 3 architecture, and the Ryzen 5000 series of processors. In addition, AMD revealed in its earnings report that the sales of Ryzen processors increased significantly in Q3.

But no matter how well AMD does, the PC space is a market with a negative outlook and little room for growth.

Since 2010, global PC shipments have been on an overall downward trend. Although the neo-crown epidemic, which required many people to purchase PCs for their home classes or offices, spurred a strong rebound in global PC shipments in Q2 and Q3 of 2020, with 9% and a 10-year high of 12.7% year-over-year growth respectively, the good times in the PC market don’t last long once demand is fully unleashed.

In fact, Intel’s Q3 2020 results show that demand for PCs has shifted from the high-margin enterprise and government markets to the low-margin personal and education markets.

In the overall downward trend in the global PC market, the CPU chip makers the wisest choice, is to turn to the data center area with more room for growth.

The share of Intel and NVIDIA’s revenue from data centers is becoming larger and larger, and in Q3 of fiscal 2020, Intel’s “data-centric” business accounted for 46% of total revenue.

In Q2 2021 (May-July 2020), for the first time, NVIDIA’s data center revenue surpassed that of its traditional gaming business – making data centers the backbone of NVIDIA’s revenue.

Looking at AMD, with its high-performance EPYC (Xodong) CPUs, AMD is taking over the data center space. Microsoft, Oracle, Amazon and Google are all using AMD CPUs in their data centers.

While AMD doesn’t report data center server revenue separately, Sozifon said at its Q2 2020 earnings meeting that they achieved double-digit share in the server processor segment for the first time.

AMD has made considerable progress in the datacenter, but it’s still CPU-dominated and short on accelerated AI computing. While AMD also has GPUs designed for the data center, they are largely non-existent compared to NVIDIA’s GPUs.

Meanwhile, competition in the data center is getting fiercer. NVIDIA’s confirmed acquisition of Arm creates a CPU+GPU portfolio that will have a direct impact on both AMD and Intel.

What does Xilinx bring to the table for AMD?

The acquisition of Xilinx, the inventor of the FPGA and the market leader in the field, will complement AMD’s acquisition of Xilinx, which Ross Freeman co-founded with colleagues in 1984 and introduced the first true FPGA chip, a Field Programmable Gate Array (FPGA).

FPGA (Field-Programmable Gate Array), is a type of hardware reconfigurable chip, the biggest advantage is the flexibility, can quickly according to the actual application scenario, the hardware level adjustment.

As Moore’s Law for CPUs enters its twilight years, and AI and cloud computing grow exponentially, data centers need to introduce new hardware to accelerate computing, such as GPUs, FPGAs and ASICs (Application Specific Integrated Circuits).

FPGAs were originally a small-scale application, but Microsoft’s massive data center deployments in recent years have allowed the chips to realize their potential: FPGAs have lower latency than GPUs. FPGAs are more flexible than ASICs and can be adapted to different neural network and computing scenarios at a lower cost.

Beyond the data center, the flexibility of FPGAs makes them ideal for edge computing. Communications, driving automation, and industrial are all scenarios in which FPGAs can be advantageous.

Currently, the FPGA market is dominated by Xilinx and Intel, which acquired Altera for $16.7 billion in 2015 and sells the latter’s FPGA products with its own CPUs to replace NVIDIA’s GPUs.

AMD’s acquisition of Xilinx also had the same synergistic effect, with the two products complementing each other. With the acquisition, AMD now has both a CPU+GPU and CPU+FPGA portfolio, giving it more ammunition in the data center market.

One interesting thing is that Dan McNamara, general manager of AMD’s server business unit, is a former vice president of Intel’s programmable business group (i.e., FPGA business). Prior to that, he worked at Altera.

M&A: A New Growth Strategy

The semiconductor industry has always been an acquisitive industry, with acquisitions of all sizes occurring every single year.Prior to 2020, 2015-2016 was the obvious peak.

Between 2010 and 2014, the total average annual value of M&A in the global semiconductor market was just $12.6 billion. 2015 saw NXP acquire Freescale for $11.8 billion, Avnet for $37 billion, Broadcom for $37 billion, Intel for $16.7 billion, and Altera for $16.7 billion… …According to research firm IC Insights, there were more than 30 semiconductor mergers and acquisitions that year, with a combined value of more than $100 billion.

The M&A boom continued in 2016, with more than 24 acquisitions totaling more than $90 billion. In the following years, the semiconductor industry’s M&A trend slowed, but still totaled $25 billion per year. Until 2019, when it started to pick up again, total M&A volume reached $31.7 billion, up 22 percent year-over-year.

The semiconductor industry will see another M&A peak in 2020. NVIDIA’s $40 billion acquisition of Arm, AMD’s $35 billion acquisition of Xilinx, Arnold’s $20.9 billion acquisition of Masonic, and SK Hynix’s $9 billion acquisition of Intel’s NAND business …… As of October of this year, M&A has exceeded the 100 billion dollars.

Clearly, the semiconductor industry is in a long period of consolidation after 2015. in January 2020, market consulting firm Accenture reported that traditional organic growth in the semiconductor industry has ended: rising R&D costs, the pace of technology iteration, and diversified customer needs have compressed the time and money that used to support organic growth in the semiconductor industry; as an alternative, leading semiconductor Manufacturers have turned to M&A as a new growth strategy, and the result has been dramatic consolidation in the industry.

Notably, as semiconductors have become increasingly important at the national level, “government regulation” has emerged as the X-factor influencing M&A in the semiconductor industry. According to Accenture, from 2013 to 2015, there were only three semiconductor M&A deals that were blocked or terminated due to factors such as “government intervention” or “regulatory restrictions”; however, by 2016 to 2018, there were only three M&A deals that were blocked or terminated due to factors such as “government intervention” or “regulatory restrictions”. That number rises to 14 cases.

One of the most well-known examples was Qualcomm’s attempt to acquire Dutch semiconductor company NXP for $44 billion in 2016. However, as China is an important market for NXP, it did not issue approvals until the two sides agreed on a deadline for the deal, and Qualcomm eventually had to abandon the acquisition.

NVIDIA’s acquisition of Arm currently has similar variables, and AMD’s acquisition of Xilinx will naturally face the same challenges.