Stock Surge
From October of last year to March of this year, the stock prices of NVIDIA and AMD both doubled, making them two of the hottest stocks in the market. Interestingly, the leaders behind these two companies, Jensen Huang and Lisa Su, are both of Chinese descent. We discussed NVIDIA last year, so today, let's talk about the other high-flyer: AMD.
Beyond Gaming GPUs
Due to the numerous discussions centered around AMD and NVIDIA graphics cards, many people assume that AMD solely focuses on gaming GPUs. In reality, AMD is not only one of the two major players in the discrete graphics card market, but also one of the two dominant forces in the CPU market. Throughout its more than half-century of business ups and downs, its perennial rival has not been NVIDIA, but Intel.
AMD's Resurgence
AMD has experienced several rises and falls in its competition with Intel. In eight years, it staged an epic comeback, with its stock price increasing over a hundredfold and its market capitalization surpassing $300 billion, instantly leaving Intel behind and even exceeding the market capitalization of Moutai. How did AMD achieve this turnaround? How did the legendary Lisa Su lead its revival and ride the wave of AI to reach its peak? Without further ado, let's discuss the remarkable story of AMD.
The Genesis of a Rivalry
The story begins with what is hailed as the greatest invention of the 20th century: the transistor. In 1947, three scientists at Bell Labs successfully created the world's first transistor, later winning the Nobel Prize in Physics for their achievement. One of these scientists was William Shockley, a master in scientific research. However, his management skills were severely lacking, plagued by issues like stubbornness, racial prejudice, and paranoia. Eight of his top disciples, themselves leading scientists, eventually grew tired of his leadership and left in 1957 to form their own company, Fairchild Semiconductor.
Fairchild's Legacy
Fairchild Semiconductor may not be a household name today, but it is no exaggeration to say it was the cradle of semiconductor talent and arguably the origin of Silicon Valley. The company continued to experience a brain drain, with all eight founders departing within a decade. The last two to leave, Robert Noyce and Gordon Moore, founded Intel in 1968. A year later, Jerry Sanders, Fairchild's sales director, founded Advanced Micro Devices, or AMD. The departure of top executives and scientists essentially signaled the end of Fairchild Semiconductor, but it also marked the beginning of the ongoing commercial rivalry between AMD and Intel.
Early Days and Intel's Dominance
Initially, AMD and Intel were not in the same league. Intel, founded by disciples of William Shockley, had both talent and capital. Investors were eager to fund them, easily raising $2.5 million. Two years later, Intel went public, continuously launching chips that set industry standards. AMD, on the other hand, did not have such a smooth start. The 1970s was a chaotic period for semiconductor startups, and small companies like AMD were a dime a dozen. No one was willing to invest. Surprisingly, the first individual investor in Sanders' venture was Robert Noyce, the head of Intel. Thus, AMD and Intel were closely intertwined from the very beginning. Sanders proved his mettle, and AMD went public just three years later, in 1972.
The Second Source Strategy
An interesting practice in the US at the time required manufacturers procuring chips, for example from Intel, to mandate that Intel share its blueprints with another manufacturer, a "second source," to prevent being locked in by a single supplier. Given their good relationship, AMD often served as Intel's second source. This strategy was similar to the "drafting" technique in cycling, where following the leader reduces wind resistance and saves energy. The leader also benefits from having someone drafting behind them. Their most successful deal came in the 1980s, securing a contract with IBM, which was entering the personal computer market, propelling both companies to success. Intel's x86 architecture became the industry standard for CPUs, and they reached their peak in the 1990s. Intel's market capitalization reached $275 billion in 2000, becoming the sixth largest company globally. AMD also saw its stock price increase 15-fold in five years.
A Turning Point and AMD's Innovation
However, as the personal computer market began to saturate and legal disputes arose over licensing, cracks appeared in the AMD-Intel relationship. AMD realized it couldn't remain tied to Intel and needed to chart its own course. In 1995, AMD significantly increased its R&D spending and acquired NexGen, a semiconductor design company, for $900 million. Interestingly, NexGen's COO was Vinod Dham, the "father of the Pentium processor," who had recently been poached from Intel. Under Dham's leadership, AMD launched its self-developed K-series architecture in 1996, ushering in a period of success. In June 1999, they introduced the Athlon processor based on the K7 architecture, quickly followed by the world's first processor with a clock speed exceeding 1 GHz, less than ten months later, directly challenging Intel's Pentium III. This processor significantly outperformed Intel's chips in terms of price-performance, generating $1 billion in profit for AMD in a single year, a stark contrast to the nearly $90 million loss the previous year.
The Acquisition of ATI and a Decade of Struggle
AMD continued its rise, launching the first 64-bit x86 processor based on the K8 architecture in 2003. By 2006, AMD's market share in the CPU market reached 48.4%, almost on par with Intel, and even surpassing Intel in the desktop segment. With momentum on its side, AMD considered its next big move: acquiring a company in the graphics card market. In the early 2000s, the CPU market was dominated by Intel and AMD, while the GPU market consisted primarily of NVIDIA and ATI. AMD believed that merging CPU and GPU technologies and customer bases would create significant synergy, potentially crippling Intel. They decided to acquire either NVIDIA or ATI. Considering NVIDIA's $7.5 billion market capitalization compared to AMD's $8.6 billion, acquiring NVIDIA risked a reverse takeover. ATI, at $5 billion, seemed a more feasible target. AMD took the plunge, exhausting its $1.8 billion in cash, borrowing $2.5 billion, and using $1.1 billion in stock to acquire ATI. This made AMD the only company with significant market share in both the CPU and GPU markets. While this seemed like a winning move, it led to a dark decade for AMD.
Challenges Mount
In the two years following the ATI acquisition, AMD wrote off $2.6 billion in goodwill, indicating that the acquisition, initially valued at $5 billion, was now deemed worth only half that amount. While AMD's strategy to enter the GPU market was sound, the acquisition price was exorbitant, straining AMD's cash flow. Intel, sensing the threat, also stepped up its game. While market share suggested parity between AMD and Intel, the reality was quite different. AMD's market share was concentrated in the low-end and mid-range segments, prioritizing price-performance, resulting in lower profit margins. Intel's revenue was roughly ten times that of AMD, and its market capitalization was thirty times larger. Intel, with its vast resources, responded to the perceived threat by restructuring, laying off employees, and refocusing its business. Just three days after AMD announced the ATI acquisition, Intel launched its Core 2 processor with a new microarchitecture, boasting superior performance and energy efficiency, widening the gap with AMD. Intel also implemented its "Tick-Tock" strategy, alternating between updating chip manufacturing processes (Tick) and processor architecture (Tock) annually. Over the next decade, Intel advanced its process to 14nm, regaining dominance in the CPU market with its i3, i5, and i7 processors.
AMD's Struggles Continue
Meanwhile, AMD was burdened by the expensive ATI acquisition, facing increasing competition from a resurgent Intel, and struggling to compete with NVIDIA in the GPU market. The strain of designing and manufacturing both CPUs and GPUs, coupled with dwindling cash flow and mounting debt reaching $5 billion, forced AMD to spin off its fabrication plants in 2008, creating GlobalFoundries. This provided some much-needed cash and allowed AMD to focus on chip design. However, AMD's product releases in the following years failed to gain traction, and sales declined from 2010 onwards. Its CPU market share plummeted from nearly 50% to 20%, and its server CPU market share fell from 30% to a mere 1%. With a market capitalization below $2 billion, AMD's survival was in question, leading to further layoffs and cost-cutting measures, including selling its headquarters for $160 million. Many Wall Street analysts predicted AMD's demise.
The Arrival of Lisa Su
But in this darkest hour, Lisa Su arrived as AMD's new CEO in 2014. A technical expert with a PhD in Electrical Engineering from MIT, Su had a strong background in chip development from her time at Texas Instruments and IBM. She inherited a company in dire straits, requiring immediate action. Su implemented classic turnaround strategies: streamlining processes, improving customer relationships, and cutting costs. However, her two key strategic shifts were crucial to AMD's salvation, both incredibly challenging to execute.
Focus on High-Performance Computing
The first shift was a focus on high-performance computing. While AMD appeared to rival Intel, its strength lay in the mid-range and low-end segments. Su recognized the limited potential of these segments and refocused AMD on developing high-performance chips. This mantra of "high-performance computing" became a constant theme in her interviews. After two years of strategic adjustments, AMD launched its Ryzen chips based on the new Zen architecture in late 2016, marking its first counteroffensive.
The Zen Architecture
The Zen architecture represented a completely new framework, different from Intel's manufacturing process. It utilized a chiplet design, integrating smaller chips into a larger one, increasing flexibility and yield. This approach became the foundation for AMD's subsequent CPUs and GPUs. While AMD's initial performance target for the new chips was a 40% improvement, considered unrealistic by many, the first-generation Ryzen exceeded expectations with a 52% improvement, outperforming Intel's comparable eight-core i7 6900K while being priced at less than half the cost. While not technically surpassing Intel at the time, continuous optimization of the Zen architecture in the following years propelled AMD forward. The 7nm process and Zen 2 architecture powered the Ryzen 3000 series in 2019, and were adopted by gaming consoles like the PS5, Xbox Series X/S, and Steam Deck. AMD's bet on 7nm paid off handsomely. The Zen 3 architecture followed in 2020 with the Ryzen 5000 series, and the 5nm process and Zen 4 architecture powered the Ryzen 7000 series in 2022. AMD's resurgence was evident not only in its chips but also in the improved quality of its product launches. The Ryzen series brought AMD back to parity with Intel in the desktop CPU market.
Expanding Business Lines: Data Centers
However, Su's ambition extended beyond consumer PCs. She recognized AMD's overreliance on the CPU market and, in 2015, identified three key growth areas: gaming, data centers, and immersive platforms. These choices proved prescient. Even before becoming CEO, Su had helped AMD by pushing its gaming business. Following the ATI acquisition, a lifeline for AMD came in 2011 with the launch of the Accelerated Processing Unit (APU), which integrated CPU and GPU onto a single chip. In 2012, Su, then in charge of products, spearheaded the adoption of AMD chips by major gaming console manufacturers, securing annual revenues of $200-300 million, offsetting losses in the PC market and keeping AMD afloat. Of Su's three target areas, AMD already dominated the console market in gaming, leaving room for steady growth. The second area, data centers, offered the greatest potential and became AMD's primary growth driver.
Conquering the Data Center Market
Data centers, used by cloud service providers and supercomputers, relied on CPUs from Intel or AMD, and GPUs from NVIDIA or AMD. The CPU segment of this market had been dominated by Intel, representing a significant portion of their business. For tech companies, cloud computing and high-performance computing were essential, even before the rise of AI. Companies often dedicated half their budgets to servers, demanding high-performance and energy-efficient chips. This was a market where AMD had previously struggled to compete with Intel. To succeed in the data center market, AMD needed to develop top-tier chips, aligning with Su's first strategic shift towards high-performance computing. The Zen architecture, coupled with Su's vision for the data center market, proved to be a winning combination. Following the launch of Zen, Su prioritized the data center business, declaring it AMD's "Number One Strategic Priority." AMD quickly launched its EPYC server chips, designed for enterprise clients. After several generations of optimization, EPYC surpassed Intel's Xeon processors in performance.
The Importance of TCO
Enterprise clients, especially tech companies running data centers, prioritized Total Cost of Ownership (TCO). Unlike consumer purchases, the initial cost of servers was a small fraction of the overall expense. Ongoing operational costs, including electricity, maintenance, and cooling, constituted the bulk of the TCO. AMD, cleverly, provided a comparison tool on its website, allowing users to compare any AMD EPYC chip with an Intel Xeon chip, analyzing price, speed, power consumption, and TCO. This transparent comparison, though aggressive, highlighted AMD's advantages in cost, performance, and energy efficiency. While AMD might have been accused of bias, Intel's lack of response suggested the validity of AMD's claims. This bold move demonstrated AMD's confidence in its products.
Gradual Gains and the Xilinx Acquisition
While AMD's EPYC chips gained traction among clients like Amazon AWS, Microsoft Azure, Google Cloud, and IBM Cloud, the shift in the data center market was gradual, due to the long-term contracts common in the enterprise sector. To further expand its reach, Su orchestrated the $49 billion acquisition of Xilinx, the leader in Field-Programmable Gate Arrays (FPGAs), a different type of chip used in diverse applications like electric vehicles, Mars rovers, and communication base stations. This acquisition, a significant undertaking for AMD with limited cash reserves, was achieved through a stock swap. Xilinx shareholders received 1.7234 AMD shares for each Xilinx share. This deal, initially valued at $35 billion, increased to $49 billion due to AMD's rising stock price during the year-long acquisition process, demonstrating the leverage gained from a strong stock performance. This acquisition marked a significant milestone in the chip industry.
The AI Boom and AMD's Continued Success
The year 2023, the year of AI, brought the fruits of AMD's previous gamble on ATI. As one of the two major discrete graphics card companies, AMD experienced a second wave of growth. Its new AI chips were immediately sought after by Microsoft, Meta, and OpenAI, boosting its data center business, which grew 43% quarter-over-quarter in Q4 2023 to $2.3 billion, surpassing gaming to become AMD's largest revenue source. AMD also integrated AI engines (NPUs) into its consumer Ryzen series, improving performance by 60%. AMD became fully committed to AI, optimizing its entire product line for AI applications.
AMD's Triumph and the Future of the Chip Industry
In just eight years, AMD's stock price increased over 100-fold, going from near bankruptcy to reaching a market capitalization of $300 billion, nearly double that of Intel, and surpassing giants like Toyota and Moutai. While Intel still leads in absolute revenue and market share, AMD’s growth potential has put it in a commanding position. The competition with NVIDIA in the GPU market, while fierce, is not a zero-sum game. Su acknowledged the vast potential of the AI market, estimated at $150 billion annually in the next three to five years, with current global AI production capacity meeting less than 20% of demand. This presents a significant opportunity for both AMD and NVIDIA, where innovation, rather than direct competition, is the key. Intel, under new leadership since 2021, has chosen a different path, investing heavily in chip manufacturing, aiming to catch up with TSMC by 2027. The battle between AMD and Intel is far from over.
The High Stakes of the Chip Industry
The chip industry is characterized by rapid innovation, where a single product can shift the balance of power. These seemingly rapid changes are the result of years of investment. Developments like AMD's EPYC and ROCm, Intel's Core processors, and NVIDIA's CUDA are the culmination of at least three to five years, and often longer, of dedicated effort. Success in the chip industry requires not only short-term product development but also long-term strategic vision, demanding constant evaluation, decision-making, and significant investment. Unlike other industries where successful bets can provide a period of dominance, the chip industry demands continuous innovation. The rapid pace of technological advancement requires companies to constantly reinvest, making it a high-stakes game of constant innovation and strategic foresight. The current winners are only temporary, and the investments being made by Intel, AMD, and NVIDIA are setting the stage for the next round of competition in this ever-evolving landscape.