Written by: Techub News Compilation

Introduction

Recently, the well-known technology podcast “All-In Podcast” held a special discussion titled “Winning the AI Race,” bringing together leading figures from key segments of the AI hardware supply chain. AMD Chairman and CEO Lisa Su, NVIDIA Founder and CEO Jensen Huang, MP Materials CEO James Litinsky, the only rare earth mining and magnet manufacturing company in the U.S., and Crusoe Energy Co-Founder and CEO Chase Lochmiller, who focuses on AI infrastructure, appeared together. Against the backdrop of intense AI competition, supply chain security, and geopolitical factors, this dialogue directly addressed the core challenges and strategic thoughts faced by the U.S. in rebuilding critical manufacturing capabilities, ensuring supply of essential materials, addressing energy bottlenecks, and cultivating the next generation workforce, providing a valuable frontline perspective for observing the global AI industry competition landscape.

Summary

• AMD CEO Lisa Su predicts that the AI accelerator chip market alone will surpass $500 billion in the coming years, and she believes it is acceptable for domestically manufactured chips to have a cost premium of “low double digits percentage” to ensure supply chain security.
• NVIDIA CEO Jensen Huang proposed the concept of “AI factories,” stating that in the future, every industrial company will have a “machine factory” to manufacture products and an “AI factory” to train AI, emphasizing America’s unique advantages in technology, innovation, and leadership.
• MP Materials CEO James Litinsky revealed the risks of U.S. dependence on China in the rare earth permanent magnet supply chain and explained in detail how its “public-private partnership” with the U.S. Department of Defense (DoD) is designed to rebuild U.S. magnet capacity through investments, price guarantees, and purchasing agreements to respond to the “physical AI” revolution.
• Crusoe CEO Chase Lochmiller pointed out that the explosive growth of AI data centers (also known as “AI factories”) is facing energy bottlenecks, predicting that their electricity consumption will surge from 2.5% of total U.S. electricity consumption to 10%, and calling for large-scale investments in energy infrastructure and workforce to support growth.
• Several CEOs emphasized that the talent shortage is a key bottleneck in rebuilding U.S. manufacturing and AI infrastructure, which needs to be addressed from multiple aspects such as education system reform, on-the-job training, and improving salary attractiveness.

Chip Manufacturing Reshoring: The Debate Over Costs, Talent, and Supply Chain Security

The discussion first focused on the real challenges of semiconductor manufacturing reshoring to the U.S. AMD's Lisa Su shared her experience in obtaining the first batch of chips (4nm process) at TSMC's Arizona plant, believing that this proves that cutting-edge manufacturing in the U.S. is feasible. When asked about the cost differences, she honestly admitted that there is a premium for U.S. manufacturing, but it is not the 50% speculated by the outside world, but rather “above 5%”, possibly in the “low double digits percentage” range (such as 10%-20%). She emphasized that in the context of today's extremely strong demand for AI chips and the critical importance of supply chain security, customers are willing to pay a certain premium for supply assurance.

However, the path to reshoring is not smooth. Lisa Su acknowledged that there are indeed initial issues such as a shortage of qualified talent, and TSMC needs to “transplant” its manufacturing experience accumulated in Taiwan to the U.S. But she expressed satisfaction with the progress, noting that the chip yield at the Arizona plant has reached parity with Taiwan. She predicts that the AI accelerator (chips for large computing systems) market will grow to over $500 billion within a few years. To meet the huge chip demand indicated by industry leaders such as Elon Musk and Sam Altman, the entire ecosystem—from chip design, manufacturing to supporting segments—needs to expand rapidly in sync.

Regarding the future of chip technology, Lisa Su believes that due to the extremely diverse AI application scenarios (from scientific computing and manufacturing to personal devices), there will be an explosion in “chip diversity” in the future, not just a few standard training or inference chips. She is particularly optimistic about the markets for “physical AI” (robots, drones, and other electric motion devices) and edge AI (local operations on personal computers, smartphones), believing that physical AI will become an “important end market” in five years or more. When asked whether AI will dominate future chip design, she stated that human creativity still plays a core role, but AI will greatly help accelerate and optimize the design process.

Rare Earth Permanent Magnets: The “Lifeblood” of Physical AI and National Security Games

If chips are the “brain” of AI, then rare earth permanent magnets are the “muscle” and key “feedstock” for “physical AI” (robots, drones, and all electric motion devices). MP Materials' James Litinsky deeply analyzed the vulnerabilities of the U.S. in this strategic supply chain. He explained that after rare earth mining, an extremely complex, expensive, and specialized chemical process of refining, separating, metallizing, and magnetizing is needed to finally produce magnets. Over the past few decades, China has dominated the global market by integrating the entire industry chain and adopting a “mercantilism” strategy—selling finished magnets at prices below raw material costs—making it difficult for competitors from other countries to survive.

Litinsky shared how MP Materials acquired the Mountain Pass mine in California from bankruptcy and gradually rebuilt U.S. rare earth mining and refining capabilities through an investment of about $1 billion, establishing a magnet factory in Texas (with General Motors as its foundational customer). However, to scale up significantly to meet anticipated demand for “physical AI” magnets in defense and commercial sectors, they still face the profound risk of price suppression from China. This is precisely the background for their “transformative” public-private partnership with the U.S. Department of Defense (DoD).

This deal is not a simple government subsidy. Litinsky detailed that the DoD gains company equity and warrants through investment, becoming an “economic investor”; simultaneously, DoD provides a “price floor” to protect MP Materials from the impacts of Chinese dumping below cost; furthermore, DoD acts as a “100% guaranteed customer,” committing to purchase all magnet output from its planned tenfold capacity factory and sharing the profits above guaranteed profit levels on a “fifty-fifty” basis with the company. Litinsky emphasized that this is a “true win-win,” safeguarding national security and critical supply chains while allowing taxpayers to potentially benefit from the market-oriented mechanism. He believes this model could serve as a “blueprint” for rebuilding U.S. capabilities in other critical areas (such as shipbuilding, advanced drug components, industrial diamonds, etc.).

Energy and Infrastructure: The “Power Hunger” of AI Factories and a Construction Frenzy

Chase Lochmiller from Crusoe Energy elevated the discussion to the macro infrastructure level. He quoted Warren Buffett’s famous saying, “Never short America,” stating that we stand at the beginning of a new era of “intelligent infrastructure.” AI data centers are evolving into “AI factories”—inputting data, algorithms, chips, and energy, outputting “intelligence.” He cited IDC data, predicting that by 2030, AI will generate $20 trillion in economic impact.

However, AI factories are “energy-hungry” beasts. The data Lochmiller presented indicated that total electricity generation and consumption in the U.S. has been relatively stable over the past 20 years, but AI is fundamentally changing this demand pattern. It is predicted that by 2030, data centers will account for 20% of U.S. electricity demand growth, with their total consumption percentage soaring from the current 2.5% to 10%. This means that if the tech industry wants to sustain AI growth, it must “bring its own energy” and make large-scale investments in energy infrastructure.

Using Crusoe’s mega-scale AI factory under construction in Abilene, Texas as an example: this facility will ultimately consume over 1.2 gigawatts of power and house 400,000 NVIDIA GPUs, forming a “gigawatt-level” single computing cluster. The project transformed from vacant land to a massive construction site within a year, with 4,000 workers on site daily, including electricians, plumbers, construction workers, and various trades, with a total investment of $15 billion. This is just the beginning; Crusoe also plans multiple gigawatt-scale projects in West Texas, Wyoming, and elsewhere, and is collaborating with companies like Redwood Materials and GE Vernova to explore the utilization of renewable energy, energy storage, natural gas, and various energy forms. Lochmiller’s core conclusion is: we need new infrastructure, large-scale infrastructure, and a large workforce to build, operate, and maintain it. AI infrastructure will become “the largest job creation catalyst in history.”

Talent, Competition, and Future Outlook: The Consensus and Vision of CEOs

No matter the topic, whether it is chip manufacturing, rare earth processing, or energy infrastructure, “talent shortage” is a recurring pain point. James Litinsky mentioned that MP Materials currently has 850 employees and needs to add “easily thousands more” as it expands for projects for Apple and the DoD. He countered the perception that manufacturing jobs are unpopular, pointing out that the median annual salary in his company is nearing $100,000, with salaries for positions such as electricians and maintenance workers reaching six figures, and employees hold company stock. Lisa Su emphasized the importance of STEM (science, technology, engineering, mathematics) education, believing that efforts need to start from early education to spark young people’s interest in science to ensure that the U.S. has the best AI talent.

NVIDIA’s Jensen Huang brought a broader perspective to the discussion. Regarding the impact of AI on employment, he sees AI as “the greatest technology equalizer in history,” enabling everyone to become programmers, artists, and writers. Although some jobs will become obsolete, many new jobs will be created. He asserted, “If you don’t use AI, you will definitely lose to those who do.” In response to Elon Musk's prediction of needing 50 million equivalent computing power of H100 in five years, Huang contextualized it within the grand backdrop of building “AI factories” and “token production” infrastructure, believing we are at the beginning of a “trillions of dollars infrastructure construction” phase.

Regarding the reshoring of U.S. manufacturing, Huang expressed strong confidence. He praised the U.S. as “the most technologically rich and innovative country in the world,” stating that the computer industry is the greatest industry the U.S. has ever created. He supported former President Trump’s vision of “reindustrializing America,” arguing that the most advanced, economy-driving, and national security-enhancing industrial segments (such as chip and AI supercomputer manufacturing) should be brought back home. He predicted that in just four years, approximately $500 billion worth of AI supercomputers could be produced in Arizona and Texas, further driving a multi-trillion dollar AI industry. He specifically pointed out that a sufficient energy supply is a prerequisite for all of this, and that current government support policies for energy and AI innovation are crucial.

Talking about the U.S.-China AI competition, Huang cited excellent Chinese open-source models like DeepSeek, pointing out two key points: first, the fortunate fact that these models operate on U.S. technology stacks, which solidifies the global position of U.S. technical standards; second, breakthroughs in “inference” efficiency allow AI to “think” for longer periods at lower costs, which is crucial for future AI development.

Finally, several CEOs envisioned an AI-driven future. Lisa Su believes that AI is the “most transformative technology of our lifetime,” with impacts on a “magnitude” level, likely to solve some of the world’s most important challenges. James Litinsky emphasized ensuring the autonomy of basic materials for the “physical AI” revolution from a supply chain security perspective. Chase Lochmiller and Jensen Huang depicted a new era supported by “AI factories” and intelligent infrastructure, characterized by unprecedented productivity improvements. Their consensus is that the outcome of this AI race depends not only on breakthroughs in algorithms but also on the solid foundation of “hard power” such as hardware supply chains, energy, and talent, and that the U.S. is fighting on multiple fronts to strengthen its leading position.

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。