Written by: Techub News Compilation

Introduction

Recently, AMD Chairman and CEO Lisa Su guest-starred on the well-known investor Reid Hoffman's podcast "Possible." In this nearly one-hour in-depth conversation, Su reflected on AMD's transformation journey from low points to brilliance, shared her personal growth experience from engineer to CEO, and delved into core topics such as the future of the semiconductor industry under the AI wave, geopolitical impacts, energy efficiency challenges, and talent cultivation. As the helmsman who led AMD's market value from the depths to heights, Su's insights on technology trends and company strategy provide a valuable frontline perspective for understanding the current AI hardware competition and global computing landscape.

Summary

• The successful transformation of AMD stems from a firm bet on "high-performance computing" a decade ago and key technical decisions, including the large-scale application of chiplets.
• AI is still in the early stage of a "super cycle," developing at a pace far exceeding any previous technology, and will become ubiquitous in the future, especially with huge potential in healthcare.
• Geopolitics and supply chain diversification are real challenges the industry must face; AMD is actively shifting some advanced manufacturing capacity to the United States, but global supply chain balance is crucial.
• Energy efficiency has become the foremost consideration in chip design; AMD has improved energy efficiency by over 30 times in the past five years and will continue to invest.
• The key to attracting and retaining top talent lies in giving them the freedom and opportunities to tackle the toughest problems, rather than merely focusing on their resumes.

From Engineer to Helmsman: A Curiosity-Driven Journey in Semiconductors

Lisa Su's career began with a simple curiosity about hardware. She recalls a childhood memory when her brother's remote-controlled car broke down; she opened it up and found a loose wire, and upon reconnecting, the car ran again. The sense of achievement from this "cause and effect" sparked her deep interest in hardware engineering. During her college years, her experience working in a semiconductor lab further fascinated her with the magical ability to build complex computing devices on a coin-sized chip.

She admits that choosing the semiconductor industry thirty years ago seemed "crazy" to many, as it was far less spotlighted than it is today. "People would ask, 'What is a semiconductor? Are you referring to potato chips?'" Su laughed. However, it was this passion for foundational technology that led her to complete her Bachelor's, Master's, and PhD in Electrical Engineering at MIT and kickstart her career at IBM.

At IBM, Su began as a device physicist working in a lab wearing a "bunny suit," gradually venturing into product management, marketing, and business operations. "At IBM, I could try something new roughly every two years." She states that this experience taught her that in the engineering field, what truly matters is your ideas and contributions, not your title or qualifications. She participated in cutting-edge projects like early copper interconnect technology, laying a solid foundation for her future leadership at AMD and guiding major technological transformations.

AMD's "Coming of Age": Betting on High-Performance Computing and Chiplet Revolution

When Su took over as AMD CEO in 2014, the company had an annual revenue of about $4 billion, with stock prices hovering around $2. Today, AMD's annual revenue has exceeded $23 billion, and its stock price has experienced a hundred-fold increase. Reflecting on this astonishing turnaround, Su believes the key decision was clarifying that AMD would "be world-class in certain areas."

"Our driving force is deciding where we can excel. I firmly believe we can be the best in high-performance computing (HPC)." She recalled that although AI had not yet exploded and the cloud computing path was different, the core team at AMD, including CTO Mark Papermaster, recognized this direction. They realized that the paradigm of Moore's Law, which had driven the semiconductor industry for thirty years, was changing, and merely relying on shrinking processes was insufficient; architectural innovation was essential.

Thus, AMD made a significant decision that would "bet the company roadmap" on fully embracing chiplet design. "At that time, I sat with our technical fellows, and we had to decide whether to stake the company's future on chiplets. We said: yes." Su confessed that this was a trust-based bet that needed validation. There were successes and setbacks throughout the process, but the team gained invaluable experience. Today, chiplet technology has become mainstream in the industry, adopted from high-performance CPUs to future AI GPUs. "It's always about making choices," she said.

Faced with the endless "trends" and noise in the tech world, how does the team focus on long-term planning over three to five years? Su's answer is faith and focus. "A company has its inherent DNA. What we really excel at is building 'big computers'—that's what we think about every day." She noted that AMD once debated whether to manufacture smartphone chips but ultimately stayed true to its core strengths. Similarly, AMD's investments in AI began seven to eight years ago, well before ChatGPT drew public attention. "What's different about large language models is that they make AI easily usable for a broader audience and application."

AI Super Cycle: Still in the "Early Innings," Healthcare is the Greatest Expectation

Su described the current stage of AI development as "early innings." Despite rapid advancements over the past two years, with AI now a common term, she believes the technology "is good but not yet excellent," and its potential remains largely untapped.

She pointed out that the speed of AI development is the fastest she has seen in her career, as the hardware iteration cycle may take three years, while software iterations can be shortened to three weeks, three days, or even less. This rapid ability to experiment and innovate is triggering huge changes at the application layer and in usage models.

When asked which global issue she most hoped AI would solve, Su unhesitatingly chose healthcare and quality of life. "You can't buy health with money, but if computing can help improve health outcomes, with AI playing a crucial role, that would be extraordinary." She painted various possibilities ranging from near to far:

• Nearby "second opinions": AI medical assistants can instantaneously provide high-quality auxiliary diagnoses, balancing service quality differences across varying regions and medical levels, especially playing a significant role in early diagnosis of diseases like cancer.
• Accelerated drug development: Shortening the drug discovery process that traditionally takes ten years of conventional computing and experimentation to a few weeks or months, greatly speeding up the pace of innovation.
• Integrating medical knowledge: The human body is an integrated system, not merely a collection of isolated organs. AI can integrate massive amounts of data and experiences, crossing specialty barriers to provide more comprehensive diagnostic insights.

At AMD, AI has also become a multiplier for productivity enhancement. Su revealed that engineers are heavily using AI tools to write code, especially in developing numerous library functions and kernel optimizations for GPUs. "The key is to make full use of hardware capabilities super simple." AMD is training AI tools to specifically optimize for its hardware, thus shortening the "lag time" between releasing new products and developers fully leveraging their performance.

Geopolitics, Energy Efficiency, and Ecology: A Complex Chessboard for the CEO

As the CEO of a global semiconductor giant, Su must navigate multiple complex issues such as geopolitics, supply chain security, and environmental challenges.

Regarding supply chains and manufacturing reshoring, she confirmed that transferring some advanced chip manufacturing capacity to the United States is "absolutely a trend that is happening" and accelerating. AMD has been actively involved for the past three to four years, with the expansion of TSMC's Arizona factory serving as a positive signal. However, she emphasized realism: "It will never be 100%. The global technology and semiconductor supply chain is highly globalized... We need to ensure we have a good balance in terms of investments and prioritization."

When addressing policies like export controls, Su stated that there needs to be a balance between protecting national security and promoting widespread adoption of American AI technology to drive innovation. "Innovation thrives when more people build in the U.S. ecosystem, as it accelerates the speed and pace of innovation." Her approach is to maintain close communication with all stakeholders, including the U.S. government, ensuring good dialogue while making clear that as a global company, AMD's goal is to "provide the best technology to the world."

Confronted with the energy challenges posed by the surging demand for AI computing power, Su emphasized that energy efficiency has become a primary consideration even more important than pure performance improvements. "We are obsessed with energy efficiency." She pointed out that if chips are more energy-efficient, people can gain more computing resources. In the past five years, AMD has improved energy efficiency by over 30 times through full-stack optimization, including enhancements in power management technology, packaging, and manufacturing processes, and plans to continue this effort in the next five years. She believes this requires collaboration across the entire ecosystem since AMD's chips are just part of the overall system.

Within the vibrant AI startup ecosystem, Su sees impressive hardware innovations. She specifically mentions that although startups often face high initial investment thresholds, the AI era is different. AMD provides infrastructure support for startups through an active venture capital program and developer cloud. She gave examples of companies like Liquid AI, which are focused on reducing the computational burden required to run powerful models, a key aspect of democratizing AI—making AI infrastructure accessible to everyone, not just giant corporations.

Talent Philosophy: Running Towards the Toughest Problems and Growing Through Challenges

Su attributes a significant part of AMD's success to talent and emphasizes the importance of "taking risks on talent." Her own career serves as an example: from a device physicist given the opportunity to manage a business and even an entire company.

"I believe that by giving people substantial and challenging tasks, they won't always succeed, but if you create an environment that allows them to truly learn in the process, you can cultivate extraordinary talent." She believes that during the period when AMD was much smaller than its competitors, the key to attracting talent was providing them opportunities to work on the most exciting technologies in the industry and learn how to achieve it.

Signs of the kind of talent she seeks include: the willingness to step up and solve problems. "My advice to people is: run towards the toughest problems. Volunteer to take on the hardest issues because you can learn the most from them." Additionally, she values talents who can integrate interdisciplinary thinking, those who do not limit themselves to being hardware or software experts but can broadly think about solutions and piece all parts together.

"School is not vocational training; it's learning how to think. Once you've learned to think, you must apply those skills every day because I'm learning new things daily." Su summarized that creating an environment that fosters continuous learning and applying new things is key to attracting and retaining top talent.

Looking Ahead: Significantly Enhancing Quality of Life with AI

At the end of the conversation, Su envisioned possibilities for the next 10 to 15 years: "I am willing to believe that in the next 10 to 15 years, we can significantly improve the quality of our lives in every aspect—health, productivity, lifespan, and more."

She believes the first step has been taken by actively applying AI across all health and medical fields, but this is not easy. "Currently, these are still two separate worlds: the tech world and the healthcare world. They are converging... We need more people with different skill backgrounds for cross-collaboration to truly bring the power of technology into daily life, making life genuinely different ten years from now." This aligns perfectly with her emphasis on interdisciplinary thinking.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。