AMD surged 1,700% in three years. What did it go through?

On July 18, 2019, AMD (Advanced Microelectronics) finally closed at a unit price of US$33.

This number is not very unusual, but when compared with the $1.83 in February 2016, you can feel how crazy it has risen. In three years, it has risen by 1,700%! This is a legendary existence in the entire semiconductor industry.

AMD's stock price trend over the past three years

What has AMD experienced in the past three years?

Before entering the skyrocketing 2016, AMD was actually in a very embarrassing position.

According to IC Insight data, in 2013, AMD's revenue was US$5.2 billion, ranking the world's 12th largest semiconductor manufacturer, surpassing MediaTek, which ranked 16th. However, its market value is only about NT$60 billion, less than one-tenth of MediaTek. Its stock price has fallen by 38% in just two months, and its revenue has declined by 45% in the past nine quarters. The survival rule of the semiconductor industry is to take advantage of the situation. At that time, its PC processors were not as good as Intel, and it could not enter the mobile phone market. Its core product graphics chips were also being left behind by NVIDIA.

At this time, Lisa Su took office and was promoted three levels to become AMD's president and CEO. She is the first female CEO in AMD's 45-year history. She was appointed at a time of defeat and was ordered to take charge of a crisis. Regarding her, an industry insider once gave this comment: "It is not an exaggeration to say that Dr. Su rebuilt AMD."

AMD CEO Lisa Su

What will her arrival bring to AMD?

In fact, within less than a week of Su Zifeng taking office, she carried out a drastic reform of AMD and proposed three major strategies: first, to create great products; second, to strengthen relationships with customers and partners; and third, to simplify operations and quickly enter the market. In her view, AMD is a company known for its innovation.

Prior to this, AMD's production line collapsed due to wrong decisions after acquiring ATi. Not only was it beaten by Intel, but it also had a long and tense competition with the graphics giant Nvidia. Competing with different powerful enemies in the CPU and GPU lines made it physically and mentally exhausted. According to relevant data, from 2007 to 2016, AMD's share in the PC chip market dropped from 23% to below 10%; in terms of server chips, AMD's market share is less than 1%. At the same time, the entire PC market shrank faster than everyone expected, and the mobile revolution that replaced the PC basically passed AMD by. AMD has been losing money for five consecutive years. When its revenue hit the bottom in 2015, it was less than $4 billion, a 39% drop from the high point in 2011. AMD is in danger.

But with the new product strategy, AMD no longer regards PC as its only source of income. Its embedded and semi-custom businesses have become its most successful and profitable businesses, making up for AMD's losses in CPU and GPU and allowing it to catch its breath a little.

At the same time, in terms of GPU, AMD established the new department Radeon Technologies Group (RTG) in September 2015 to develop more competitive GPU products. Technical expert Raj Koduri became the business director of RTG and created the Polaris family of GPUs that became popular later. The Polaris architecture uses the fourth-generation GCN core and FinFET process. The energy efficiency ratio is very good under the 14nm process, which successfully helped AMD increase its independent graphics market share from 20% to 30%, and AMD's stock price also soared.

According to Baidu Encyclopedia, on January 5, 2017, AMD announced the preliminary details of the Radeon VEGA GPU architecture. The Radeon VEGA architecture has been planned and designed for more than 5 years. Compared with traditional GPU architectures, Radeon VEGA opens up new possibilities for PC gaming, professional design, and machine intelligence. The VEGA GPU architecture is an important and key product for AMD to re-enter the high-end graphics card market.

This year, AMD processors officially entered the 7nm era. At AMD's shareholders meeting in May, Su Zifeng mentioned the following sentence: "In the field of gaming, we launched the Radeon VII GPU in February and successfully returned to the high-end market. We are expected to launch the next-generation Navi GPU for the mainstream market later this year", indicating that in the second half of this year, AMD will launch a GPU based on the new 7nm process Navi architecture.

As for CPU, AMD is also preparing a product that will allow it to reappear in front of the public. This product must have the advantage of competing with its old rival Intel. Before this, AMD had thought about taking a differentiated route. In order to compete with Intel in the high-end market, it planned to start with multi-core, cost-effectiveness, integrated platform and other aspects. Therefore, in 2011, AMD once created a desktop processor called "Bulldozer", focusing on the core graphics part, which hit Intel's weak GPU technology, but ultimately failed.

This time they regrouped and prepared to develop a new architecture - Zen. In their expectation, this new processor architecture must be competitive enough in the high-end desktop market and at least 40% higher than the IPC of the excavator. The Zen architecture is a CPU architecture that has been completely redesigned from the bottom up, with equal emphasis on performance and energy efficiency. Its single-threaded, multi-threaded, and energy efficiency are comparable to Intel Skylake/Kaby Lake. Its basic component is the core complex (CCX): four cores as a unit, running eight threads at the same time, which just confirms AMD's belief in multi-core and multi-threaded in the design of desktop processors. It will be widely used in many markets such as high-performance desktops, servers, notebooks, and embedded systems, and is planned to be first used in high-performance desktop processors code-named "Summit Ridge".

At the time, Intel's latest processor was Kaby Lake, but Kaby Lake was only available in dual-core and quad-core form factors, some with simultaneous multithreading (SMT) and some without. If you wanted more than quad cores you had to go back to the previous generation processor architecture - Broadwell.

For multi-threaded intensive work, Zen and Broadwell are comparable in performance; for work that relies on AVX or memory bandwidth, Zen has an overwhelming advantage. Of course, most importantly, compared with Intel's $1,049 chip, AMD only costs $499, which fully demonstrates the value of cheapness.

In 2017, the new Ryzen series CPUs with Zen architecture helped AMD regain its position as an industry leader, ending the decade-long era of four cores and eight threads for mainstream CPUs. Although Ryzen's main frequency is not as good as Core i7, its eight cores with high cost performance have helped Intel to upgrade its mainstream CPUs from four cores and eight threads to six cores and twelve threads, and then to eight cores and sixteen threads, making users realize that eight-core processors can also be used at mainstream prices. AMD can finally raise its head and compete with Intel. For this reason, Su Zifeng has publicly stated that this will put AMD back on the same starting line as Intel in the high-end processor market.

Based on the new Zen architecture, AMD also launched the EPYC series of processors in the same year, competing with Intel's Xeon series processors for the server market. At that time, Intel occupied 99% of the market share in this field worth $17.2 billion. In the promotional video when AMD Snapdragon was released, AMD directly confronted Intel: "People used to think that Intel Xeon was the best server solution. Now, it's time to reconsider this issue."

When the first generation of EPYC was released, it was ridiculed as "glue multi-core". Intel believed that despite its large number of cores, the multi-cores would hinder each other and greatly affect the overall performance. Most media also believed that AMD still had a long way to go to catch up with Intel Xeon in a short time. However, in just one year, the breakthrough of 7nm process and the emergence of Zen2 architecture made this possibility a reality. It can be said that the emergence of the first generation of EPYC in 2017 marked the official return of AMD to the server market.

For AMD, 2017 is a big year for products, which fits Su Zifeng's strategy of shortening operations and quickly entering the market, so as to occupy the terminal market at different price points by releasing different products. Because AMD knows that there is no time to delay, and ten years of rest are only for today's rapid rise.

Entering 2018, given that Intel's 10nm process, which was supposed to be launched in 2015, has been delayed for 3 years, the industry's comments on Moore's Law have begun to increase. At this time, AMD's old friend GlobalFoundries announced that the development of the 7nm process would be postponed indefinitely, and withdrew from the competition for advanced processes, which caused an uproar for a while, and also troubled AMD, because the company's new Zen2 processors and Vega/Navi graphics card GPU chips will use 7nm technology. Fortunately, AMD responded quickly and turned to TSMC. In fact, when the second-generation Ryzen was produced, AMD had already been riding two boats and used two foundries at the same time, so making such a decision was expected. GlobalFoundries' process level has always been slightly inferior to TSMC and Samsung. This time AMD's switch to TSMC will make its product performance even more powerful.

When analyzing something, we cannot just look at the thing itself, just as in literary history, the evaluation of a person must be placed in the context of the times. The rapid development of AMD in recent years is driven not only by its own reasons but also by just the right external factors.