Photo credit: www.cnbc.com
Lisa Su, the CEO of Advanced Micro Devices (AMD), recently attended the Artificial Intelligence Action Summit at the Grand Palais in Paris on February 10, 2025.
When Su took the reins at AMD in late 2014, the company faced significant financial challenges and was perilously close to bankruptcy. At that time, AMD struggled to compete in the crucial computer processor market dominated by Intel. Amidst billions in debt and commitments to produce chips for uncertain clients, AMD’s market value plummeted to about $2 billion, in stark contrast to Intel’s valuation of approximately $180 billion.
Reflecting on her decision to lead AMD, Su noted in a recent Stanford University speech that many mentors cautioned her against taking the position, doubting its potential for positive change.
Today, Su finds herself in an entirely different situation. AMD surpassed Intel in market valuation in 2022, with its worth skyrocketing to around $172 billion—an impressive 85-fold increase since she became CEO. The company’s processors now power millions of gaming consoles, including the Microsoft Xbox and Sony PlayStation, and the U.S. government has classified AMD chips as essential to national security.
Despite these victories, AMD still perceives itself as an underdog, particularly in the realm of artificial intelligence (AI), where it significantly trails Nvidia, a market leader valued close to $3 trillion in the graphics processing unit (GPU) sector. With major tech companies investing billions in advanced AI infrastructure, AMD recognizes the need to capture a larger portion of the AI GPU market.
Su’s approach hinges on her engineering expertise, emphasizing that the semiconductor industry thrives on making strategic technical decisions that yield high-performance chips. She believes the ripple effects of today’s choices may take three to five years to materialize in product offerings.
“One of the things that I like to say about the semiconductor industry, or technology in general, is that the decisions we make today have impacts that we will recognize three to five years down the road,” Su stated during her Stanford address, underscoring the importance of making informed bets.
In an industry dominated by men, Su stands out not only as the sole female CEO among the top ten semiconductor firms by market capitalization, but also as the highest-paid woman in her position for five consecutive years, according to an Associated Press survey. The broader corporate landscape, however, still shows a disparity—only about 10% of Fortune 500 CEOs are women, with the number increasing to 52 in 2024 from 41 in 2021.
Taking calculated risks
Su, who has a penchant for strategic risk-taking in her personal life, such as her fondness for Texas Hold ’em poker, is making a significant bet on the future of AI in her professional capacity. AMD plans to unveil new AI chips annually and is establishing a software division composed of thousands of employees to develop open-source tools that can rival Nvidia’s dominant CUDA software.
As AMD’s technology chief Mark Papermaster noted, Su’s active involvement in the company’s technology strategy is vital, as she frequently visits chip labs and engages with engineering teams.
“Lisa is greatly respected by the engineering team because she understands their work, and they appreciate her expertise,” Papermaster explained.
Born in Taiwan in 1969, Su moved to the United States as a child when her father pursued graduate studies. She has characterized her upbringing as a “typical Asian upbringing,” emphasizing academic achievement, and later attended The Bronx High School of Science in New York.
Su’s introduction to the semiconductor field occurred during her undergraduate studies at the Massachusetts Institute of Technology (MIT), where she later earned a Ph.D. in electrical engineering. Her groundbreaking research on silicon-on-insulator technology in 1992 significantly impacted chip manufacturing in the subsequent decades.
After completing her education, she worked at Texas Instruments and then at IBM, where she managed research and design teams focused on chip manufacturing and served as a technical assistant to IBM’s chair and CEO at the time.
Su later became the Chief Technology Officer at Freescale Semiconductor before joining AMD in 2012, tasked with managing the business units. Her promotion to CEO in 2014 marked the beginning of a remarkable turnaround for AMD, which was then struggling to remain viable.
“AMD had serious cash flow issues,” remarked Patrick Moorhead, a former AMD executive. “They had to completely reset the company.”
On her first day as CEO, AMD’s stock traded at just over $3 per share, whereas today, the share price exceeds $106. The company is now the fourth-most valuable chipmaker in the U.S., ranking behind Nvidia, Broadcom, and Qualcomm.
Immediately upon taking the helm, Su made tough decisions, including implementing staff cuts totaling 7% of the workforce. She focused on developing partnerships with Sony and Microsoft to supply chips for their game consoles, aimed at improving cash flow and efficiently utilizing the company’s manufacturing facilities.
During her talk, Su emphasized that it wasn’t simply cost-cutting measures that rejuvenated AMD. Instead, the investment in new technology has proven crucial.
“In technology, there’s no victory in cutting yourself to become a winner,” she asserted.
Su redirected AMD’s focus towards innovation, encouraging teams to prioritize creating high-performance chips. She also chose to halt sales of certain server chips until AMD could develop offerings competitive with Intel’s performance standards, as noted by Papermaster.
Revolutionizing chip design
Aging processor designs posed a significant challenge for AMD. Su championed a comprehensive redesign of AMD’s compute core, a critical asset for the company. This initiative took several years, and the market first saw products based on the new “Zen” architecture in 2017, which has now undergone five generations of enhancements.
Su was also pioneering in the adoption of “chiplets,” a technology allowing AMD to assemble smaller chips instead of fabricating one large chip with all necessary components. This innovative approach increased manufacturing flexibility and efficiency and mitigated risks associated with production, particularly given AMD’s constrained cash flow. The strategy aligned well with AMD’s shift to utilizing Taiwan Semiconductor Manufacturing for chip production in 2019.
Emphasizing chiplets played a significant role in AMD’s growth as Intel dealt with challenges related to its investments in less effective manufacturing technologies, according to Scott Thompson, a professor of electrical engineering who has followed Su’s career for years.
“Lisa Su fundamentally understands chip manufacturing,” Thompson remarked. “She has the capability to evaluate risks and select the optimal technological pathways.”
Chiplets are now ubiquitous among leading processor manufacturers and served as a key component in AMD’s inaugural major GPU release. The proficiency developed in chiplets enabled AMD to adapt a data center GPU, initially designed for supercomputing, to become suitable for AI by substituting the CPU chiplets.
AMD’s sizable GPU, the Instinct MI300X, was launched in 2023, emerging as the demand for generative AI surged following the release of ChatGPT.
Looking toward the future
AMD experienced a 14% increase in revenue in 2024, reaching approximately $26 billion—nearly quintuple its earnings when Su began her tenure. The company dedicated $6.5 billion to research last year, more than six times its investment in 2014.
However, AMD’s position in the GPU market still mirrors its previous competition with Intel in CPUs, as it seeks to establish itself against Nvidia. While AMD has made notable advancements with AI accelerators and recorded $5 billion in AI chip sales last year—up from $100 million in 2023—there remains a significant gap compared to Nvidia’s $115 billion in data center chip and networking sales during the same fiscal period.
AMD’s stock has seen a decline of over 10% in 2024 following an 18% drop the previous year, a reversal from a 128% surge the year before.
One of the most pressing challenges for Su lies in the software arena. Despite AMD’s hardware capabilities rivaling those of Nvidia and success in certain benchmarks, many AI developers are entrenched with Nvidia’s proprietary CUDA software, which facilitates access to critical GPU functionalities for AI endeavors.
In response, AMD is focusing on ROCm, its alternative to CUDA that is available free of charge. Su aims for ROCm to attract developers seeking options beyond Nvidia’s platform.
AMD recently restructured parts of its software teams, including those working on ROCm, to form a new AI software division. The company has also started open-sourcing essential components, enabling developers to utilize AMD’s AI chips more effectively.
To foster adoption of ROCm, Su understands the necessity of improving AMD’s outreach to AI developers and shifting away from a perception of being merely a hardware provider.
In December, Su directly addressed a public critique of AMD’s software on social media, stating that the company had invested heavily in optimizing for customer needs while acknowledging there was more work ahead.
While Su tends to downplay the competitive stakes with Nvidia, she envisions a substantial growth trajectory for the AI chip market, anticipating it could expand to $500 billion annually by 2028—equivalent to the pre-AI boom semiconductor market size.
She is confident that AMD is well-positioned to capture a significant share of this emerging market.
Don’t miss these insights from CNBC PRO
Source
www.cnbc.com