AMD, Nvidia chip acquisitions target Intel's dominance

AMD, Nvidia and Intel raised the competitive heat in the chip industry with all three making strategic acquisitions to run AI workloads.

Faced with the classic decision of build versus buy to grow their companies, the major chipmakers have decided to buy. 

Intel, AMD and Nvidia recently opened up their wallets to acquire other chip companies that complement each of their respective strengths on the hardware side over the short term and their software strategies over the longer term.

Last week AMD, long the second choice to Intel in the x86 market, made a bold move with the acquisition of Xilinx in an attempt to become the leader in high-performance computing. The whopping $35 billion stock-based transaction would bring under one roof a portfolio that has a blend of CPUs and GPUs, along with field programmable gate arrays (FPGAs) and system on a chip (SoC) aimed to boost the company's fortunes in a number of markets, most notably cloud and the edge.

Analysts said the acquisition gives AMD expanded opportunities in the commercial and industrial markets that place a strong emphasis on high performance computing, given the plans of many enterprises to implement AI software that will run across multi-cloud environments.

"AMD's silicon solutions have gained traction [under AMD CEO Lisa Su], but the company is a fraction of the size of chief rival Intel and tends to be an also-ran in the enterprise and data center markets," said Charles King, president of Pund-IT, an IT industry analysis firm. "Xilinx could help fix that situation and make AMD less dependent on the often-variable PC and gaming markets."

AMD has invested in software and hardware development to compete in the cloud and edge markets, but it hasn't been as aggressive as its more cash-rich rivals Nvidia and Intel. How much of this weighs on its competitive fortunes moving forward will be determined over the next year or two.

"I'm sure AMD understands the value [of AI], but the company simply doesn't have the resources Intel has for such development efforts at this point," King said.

Still, AMD's focus on raw compute power is paying off. Last month the company released its Ryzen 5000 Series processor for desktops. It has 16 cores and is designed to handle data-intensive workloads. In the third quarter of 2020, Intel's share of the x86 market fell to 62.6% from the 64.9% in the previous quarter, according to a report released by Statistica, while AMD share rose to 37.3%. However, Intel held a commanding lead in the laptop CPU market in the third quarter with an 80% with AMD trailing far behind at a 22.4%.

Adding even more processing firepower, AMD is set to release its Radeon RX 6000 Series graphics cards that use its Navi 21 GPU, code-named "Big Navi," a chip it hopes will loosen the stranglehold Nvidia has on the GPU market. The new AMD graphics card reportedly can outperform Nvidia's GeForce RTX 2080 Ti.

Despite Intel's sizable market share lead, some analysts believe AMD's Xilinx acquisition and the upcoming GPU offering will make the company a more serious competitor to both Intel and Nvidia.

"The Xilinx acquisition shows AMD is getting more confident to compete in the enterprise server area against Intel," said Nicholas Grizzell, a research analyst with Nucleus Research. "This can put [AMD] on a much more solid footprint in the enterprise server area, especially since Nvidia is going to be in that area competing against Intel as well."

AMD over the past five years has moved from being a second-tier player to become a force in the desktop, notebook and data center server market, said Patrick Moorhead, president and principal analyst with Moor Insights & Strategy. He cautioned, however, that it typically takes five years before a brand-new architecture becomes mainstream.

"AMD has made strides on the GPU front, but not as much with CPUs," Moorhead said. "It looks like it just fielded a very competitive consumer GPU with 'Big Navi,' but we don't know yet how that translates to the data center training or inference markets."

While most IT buyers use Intel-based x86 products now, they will have more choices in the future for new workloads such as AI in the aftermath of the AMD-Xylinx and Nvidia-Arm acquisitions, said Alan Priestley, a vice president at Gartner who advises clients on the semiconductor market.

"The volume of data is changing that IT organizations have to manage and process, and to do that, they need some of these new technologies," Priestley said. "They need GPUs. In the future, they're going to end up using bigger FPGAs. They're going to end up using dedicated chips for these workloads."

Intel acquisitions target AI

Intel has also been busy counterpunching its archrivals. The company this week acquired, an Israeli-based company that has a platform for data scientists to build and run machine learning models. It also recently bought SigOpt, which has a platform for optimizing AI software models at scale. And while AMD and Nvidia are poised to gain ground on Intel, the chip giant has revenues of $72 billion compared to AMD's $6.7 billion and Nvidia's $10.9 billion.

However, Intel this year failed to bring its 7-nanometer chip to market due to issues involving its manufacturing process -- something its rival AMD delivered in July 2019 and Nvidia in November of that year. While analysts are hardly calling for the demise of Intel, they say delays in delivery of key offerings and poor management over the past few years have put the company at a significant competitive disadvantage.

Intel seems to always come back from bad news a year later. But each time they've come back the last couple of years the improvements are marginally lower and things are not as innovative.
Nicholas GrizzellResearch Analyst, Nucleus Research

"Intel seems to always come back from bad news a year later," Grizzell said. "But each time they've come back the last couple of years, the improvements are marginally lower and things are not as innovative."

Not only is Intel under attack from technologically aggressive competitors, time is running out on a technology law the company itself invented -- Moore's law. And because Intel hasn't come out with enough compelling add-on technology to compensate for the increasing limitations on x86 chip performance, the company faces even greater challenges.

Intel told one analyst it is trying to refocus its business away from its central processing-oriented business to focus on playing a bigger role across a much wider range of different applications.

"When you have as dominant a lock on the chip market as they had, and then you have a rapid proliferation of specialty hardware-assist devices that need different manufacturing, you have nowhere to go," said Geoffrey Woollacott, senior strategy consultant and principal analyst at Technology Business Research. "They have put themselves behind Nvidia and AMD by a couple of years," he said.

Although Intel has stumbled, Priestley does not think the chip giant is in serious trouble. But he does see other vendors gaining share in the market.

"What we'll see is that Intel will probably diversify its product portfolio so it keeps its business model, but it may no longer be the dominant supplier in the marketplace," Priestley said.

When Intel struck an $9 billion deal with SK Hynix last month to sell off its NAND flash memory and storage business, the chip giant said it planned to focus on long-term priorities in AI, 5G networking and edge computing as well as its distinctive Optane memory products. Intel recently partnered with VMware to provide 5G infrastructure, and it is building out AI-focused processors on technology it acquired from Habana Labs in December for $2 billion.

Nvidia spends big to gain ground

With its $40 billion purchase of Arm in September, Nvidia made perhaps the strongest move to improve its position in the overall chip market. The company has long been the leader in the GPU market, partnering with the majority of top-tier server makers and maintaining a laser focus on AI.

With its acquisition of Arm, Nvidia gains entry to a number of new markets, from supercomputers to cell phones and IoT devices. Arm processors are found in the majority of the world's smart phones, and in July, Riken and Fujitsu announced they had co-developed the world's fastest supercomputer based on Arm processors and surpassed IBM's Summit supercomputer.

Adding more momentum to Arm, Apple plans to replace Intel chips with Arm processors in its next generation Mac systems.

During the past few weeks, however, questions have surfaced about the deal as it faces significant regulatory approval hurdles in several countries. Critics of the deal cite concerns about limiting competition and favoring Arm's new owner.

"Of the three acquisitions by the three companies [AMD, Intel, Nvidia], I can't help thinking that Nvidia's pursuit of Arm is in serious trouble," said one consultant. "The complaints from Arm's founders need to be taken seriously, but issues in China could be more problematic, especially if the government there decides it needs to show some muscle," he said.

Next Steps

Pat Gelsinger reorganizes Intel, creates data center, AI group

Nvidia and Intel partner to power new workstations

Dig Deeper on Data center hardware and strategy

Cloud Computing
and ESG