The Thinking Machine

1. Resilience Forged in Adversity

「那時沒有輔導人員，」黃仁勳說：「沒人會聽你訴苦，你只能堅強起來，繼續前進。」

Early life challenges. Jensen Huang's childhood was marked by significant adversity. Sent from Taiwan to rural Kentucky at age 10, he faced bullying, racial slurs, and lived in a boarding school that was effectively a reformatory. Sharing a room with a 17-year-old illiterate roommate with knife scars, he learned to teach reading and do push-ups, developing a toughness that would define his career. This period instilled in him a self-reliance and resilience crucial for navigating the volatile tech industry.

Adapting to new environments. Despite the difficult circumstances, Huang excelled academically and socially. He quickly adapted to American culture, learned English, and found passions in computer science and table tennis. His ability to thrive in challenging, unfamiliar environments, whether crossing a dilapidated bridge to school or navigating the complexities of the tech world, became a hallmark of his approach. This early experience taught him the value of perseverance and the ability to find opportunity even in the most unpromising situations.

Finding strength in hardship. Huang's perspective on this period is notably pragmatic, viewing it not as trauma but as a formative experience that taught him the value of hard work and self-sufficiency. He learned to rely on himself when no one else was available to help, a lesson that would serve him well when making high-stakes decisions later in life. This ability to frame hardship as a source of strength is a key aspect of his enduring drive.

2. Mastering the Fundamentals of Silicon

「他有能力讓1＋1＝3，」霍茨曼說。「我的意思是，我們不只為客戶做出他們想要的東西，還把這些訂單轉化為工具，再把這些工具變成一套標準的操作方法。」

Foundational engineering skills. Huang's career began at AMD and LSI Logic, where he honed his skills in electrical engineering and chip design. Working with tools like SPICE, he learned to manipulate circuits at a fundamental level, pushing performance beyond perceived limits. This deep understanding of how silicon works from the ground up became a critical advantage throughout his career.

Going beyond expectations. At LSI Logic, Huang was known for his ability to not just meet client demands but to exceed them, turning custom projects into standardized methodologies. This knack for extracting maximum value and creating reusable tools from individual tasks demonstrated an early strategic mindset, seeing beyond the immediate problem to build foundational capabilities. His colleagues recognized this ability to make "1+1=3" in engineering.

Continuous learning and adaptation. Despite starting with paper designs and colored cellophane, Huang embraced new technologies like VLSI design tools. He pursued a Master's degree at Stanford while working full-time, constantly updating his knowledge in a rapidly evolving field. This commitment to continuous learning, even as he rose through the ranks, ensured he remained at the cutting edge of chip design.

3. Embracing Failure as a Catalyst

「我們錯過了一切，」黃仁勳談到這段草創歲月時說道：「我們做的每一個決定都是錯的。」

The NV1 disaster. Nvidia's first product, the NV1 graphics chip, was a commercial failure. Despite initial promise and a deal with Sega, it relied on a non-standard rendering method (quadric textures) that the market rejected, especially after Microsoft introduced the DirectX standard favoring triangles. This led to massive returns, near bankruptcy, and the painful layoff of most employees.

Learning from mistakes. Instead of being defeated, Huang viewed the NV1 failure as a crucial learning experience. He ruthlessly analyzed what went wrong, concluding that they had made every possible mistake. This led to a radical pivot, abandoning their proprietary architecture to align with Microsoft's standard and focusing on speed and cost-effectiveness.

The "Thirty Days to Live" mantra. The near-death experience of the NV1 failure became a core part of Nvidia's culture, embodied in Huang's mantra, "We are thirty days from going out of business." This constant reminder of existential threat fostered a sense of urgency, resilience, and a willingness to take calculated risks, ensuring the company never became complacent, even during periods of immense success.

4. The Audacious Bet on Parallel Computing

「矽谷遍地是平行運算平臺公司的屍骸，」他說：「除了我們，沒有一間採用平行運算的公司成功過，一間都沒有，前人的成功率就是零。」

Identifying a hidden opportunity. While developing graphics chips for video games like Quake, Nvidia engineers realized their chips' architecture, designed to process many pixels simultaneously, was inherently suited for parallel computing. This was a stark contrast to traditional CPUs that processed tasks sequentially. Despite parallel computing's history of commercial failure, Huang saw its potential.

Defying conventional wisdom. Huang's decision to invest heavily in making their GPUs programmable for general-purpose parallel computing (GPGPU) was highly contrarian. The market was small, programming was difficult, and industry giants like Intel were focused on improving sequential CPU performance based on Moore's Law. Wall Street saw it as a waste of resources, calling it the "CUDA tax."

The power of the GPU. The GPU's ability to perform thousands of simple calculations simultaneously made it vastly superior to CPUs for tasks that could be broken down into parallel processes. While initially applied to graphics rendering, Huang envisioned its use in scientific computing, simulations, and eventually, AI. This bold bet, made when the market was non-existent, positioned Nvidia for future dominance.

5. Building an Ecosystem Around CUDA

「自此，你就再也不想離開了，」阿爾茨說：「這就是供應商鎖定（vendor lock）。把你綁得死死的，無法走出去。」

The CUDA platform. Recognizing the difficulty of programming parallel hardware, Nvidia invested heavily in developing CUDA, a software platform that made it easier for developers to access the GPU's parallel processing power. Launched in 2006, CUDA was initially met with skepticism and low adoption rates.

Cultivating the "zero-billion-dollar market". Huang strategically targeted academic researchers and niche scientific fields with CUDA, a market he called "zero-billion-dollar." By providing free software tools and support to scientists working on problems like climate modeling or medical imaging, Nvidia built a loyal base of users who became dependent on the CUDA ecosystem. This was a long-term play, sacrificing short-term profit for future market creation.

Vendor lock-in through software. The true genius of CUDA was creating a powerful software moat around Nvidia's hardware. While competitors could potentially replicate Nvidia's chip designs, they lacked the decade-plus head start in building a comprehensive, user-friendly software stack. This created a powerful vendor lock-in, making it incredibly difficult and costly for researchers and companies to switch to competing hardware.

6. Recognizing the AI Inflection Point

「他立即看出潛力，比所有人都早一步，」柯克說：「他領先群倫，看清AI可能的發展。他真的就是第一人。」

The AlexNet breakthrough. In 2012, researchers at the University of Toronto, using two off-the-shelf Nvidia GPUs and the CUDA platform, achieved a dramatic breakthrough in image recognition with their AlexNet neural network. This demonstrated that deep learning, a long-dormant field, could achieve unprecedented results when powered by parallel processing.

Huang's rapid pivot to AI. While the AI research community was initially slow to grasp the full implications of AlexNet, Huang immediately recognized it as the killer application for Nvidia's parallel computing architecture. He swiftly directed the entire company's focus towards deep learning, declaring Nvidia an "AI company" almost overnight. This rapid, decisive pivot was crucial to seizing the emerging opportunity.

The insatiable demand for compute. Huang understood that training and running large neural networks required an exponentially increasing amount of computational power, a demand that only Nvidia's GPUs, with their parallel architecture and CUDA software, could efficiently meet. He saw that AI's hunger for compute would dwarf the demands of graphics or scientific simulations, creating a massive, enduring market for Nvidia.

7. Scaling Intelligence with Transformers

「我們看到的是，隨著它的規模愈來愈大，它顯然也好像愈來愈聰明！」薩澤爾說：「我們以前寫出的程式可不是這樣。」

The Transformer architecture. In 2017, Google researchers developed the Transformer architecture, a novel neural network design that revolutionized natural language processing. Unlike previous recurrent neural networks, Transformers could process entire sequences of text in parallel, making them highly efficient on GPUs and enabling the creation of much larger language models.

Emergence of large language models (LLMs). The combination of the Transformer architecture, massive datasets, and the availability of powerful Nvidia GPUs led to the development of Large Language Models (LLMs) like OpenAI's GPT series. These models exhibited surprising "emergent" abilities as they scaled, performing tasks they weren't explicitly trained for, such as writing code or passing exams.

The AI factory for language. Training and running these massive LLMs required unprecedented computational resources, solidifying Nvidia's position as the essential provider of AI infrastructure. Companies like OpenAI and Google became Nvidia's biggest customers, building "AI factories" – vast data centers filled with tens of thousands of Nvidia GPUs – to power the language revolution.

8. The Paradoxical Leadership of Jensen Huang

「你覺得自己不能讓他失望，」克雷說：「你就是不能。」

Demanding and inspiring. Jensen Huang is known for his intense, demanding, and sometimes volatile leadership style. He is famous for publicly berating employees who make mistakes, creating a culture of both fear and intense loyalty. Employees describe feeling immense pressure not to disappoint him, driving them to achieve seemingly impossible goals.

Loyalty and care. Despite his harsh critiques, Huang fosters deep loyalty among his employees. He is known for remembering details about their lives, providing support during personal crises, and rarely firing people for performance issues, preferring to use public criticism as a teaching tool. This blend of demanding expectations and genuine care creates a unique, high-pressure, yet supportive environment.

Visionary and detail-oriented. Huang possesses a rare combination of long-term vision and obsessive attention to detail. He can articulate a revolutionary future for computing while simultaneously scrutinizing the placement of a water cooler or the stitching on a cartoon character's pants. This ability to operate at both the highest strategic level and the most granular operational level is key to Nvidia's execution speed.

9. The AI Factory and Infrastructure Dominance

「一個新的產業正在開始：由AI資料中心處理大量原始數據，提煉為數位智慧，」他說。「就像是上一次工業革命的交流電廠；輝達的AI超級電腦基本上就是這次工業革命中的AI生產工廠。」

Building the AI infrastructure. Recognizing that AI required a new type of computing infrastructure, Huang spearheaded the development of integrated hardware and software solutions specifically for AI workloads. This included the DGX systems, which package multiple GPUs into powerful AI supercomputers, and the acquisition of Mellanox to provide high-speed networking.

The "AI factory" concept. Huang frames modern data centers as "AI factories" that consume raw data and produce intelligence. This analogy highlights the industrial scale and transformative output of these facilities, positioning Nvidia's GPUs as the essential machinery of this new industrial revolution, much like power plants were for the last one.

Software as a moat. Beyond hardware, Nvidia's extensive software ecosystem, including CUDA and hundreds of domain-specific libraries, is crucial to its dominance. This software stack makes it vastly easier to develop and deploy AI applications on Nvidia hardware, creating a powerful barrier to entry for competitors and ensuring continued demand for Nvidia's chips.

10. Geopolitics and the Silicon Shield

「一旦臺灣和台積電出事，影響之大，就像是要我回答如果加州掉進海裡要怎麼辦一樣，」黛博拉．蕭奎斯特這麼說。

Dependence on TSMC. Nvidia's business model relies heavily on TSMC in Taiwan for manufacturing its advanced chips. TSMC's unparalleled precision and scale are critical to Nvidia's ability to produce cutting-edge GPUs. This dependence links Nvidia's fate directly to the geopolitical stability of Taiwan.

Taiwan's "Silicon Shield". The concentration of advanced semiconductor manufacturing in Taiwan, particularly at TSMC, has led to the concept of a "silicon shield," suggesting that the world's reliance on Taiwanese chips deters potential aggression. Huang, despite his Taiwanese heritage, publicly downplays the risk, focusing instead on the strength of the partnership.

Global competition and supply chain shifts. Geopolitical tensions, particularly between the US and China, are driving efforts to build semiconductor manufacturing capabilities in other regions like the US, Europe, and Japan. While this diversification aims to reduce reliance on Taiwan, the scale and complexity of replicating TSMC's most advanced processes remain immense, ensuring Taiwan's critical role for the foreseeable future.

11. Unprecedented Wealth and Future Challenges

「我擁有的錢比我想像的多太多了，」他說。

Explosive financial growth. Nvidia's stock price has seen astronomical growth, particularly since the AI boom began. This has made Jensen Huang and many long-time employees billionaires and multimillionaires, generating wealth on a scale rarely seen in corporate history. The company's market capitalization has soared, making it one of the most valuable companies globally.

Market dominance and scrutiny. Nvidia holds a near-monopoly in the AI chip market, leading to massive profits and high margins. This dominance attracts both admiration and scrutiny, with accusations of price gouging and concerns about the concentration of power in one company. Competitors are actively trying to break Nvidia's hold, but the strength of its ecosystem makes this a formidable challenge.

Future uncertainties. Despite its current success, Nvidia faces significant challenges. The sustainability of the AI boom, the immense power demands of AI factories, geopolitical risks related to Taiwan and China, and increasing competition all pose potential threats. The company's value is heavily based on future expectations, creating pressure to continuously innovate and deliver in a rapidly evolving landscape.

Last updated: July 9, 2025