Sacks Sees Case for Selling China ‘Deprecated’ AI Chips

Key Concepts:

• AI Boom
• Market Competition
• U.S. Economic Growth
• Pro-Innovation Policies
• Infrastructure & Energy for AI
• "China Hawk" Stance
• U.S.-China AI Race
• Global Market Share
• American Technology Stack
• Export Control (Deprecated Chips)
• Talent Pipeline
• Ecosystem Dominance
• "Behind the Meter" Power Generation
• GPUs

---

The AI Boom and Market Dynamics

The discussion highlights an ongoing "AI boom" that is driving businesses to unprecedented highs. This boom is characterized by intense competition and cooperation among leading model companies and chip manufacturers. The collaboration between a U.S. chipmaker like AMD and a "frontier model shop" like OpenAI on infrastructure is cited as a prime example of this dynamic. This growth is attributed to President Trump's "pro-innovation, pro-export, pro-AI policy," which reportedly contributed to a 3.8% U.S. GDP growth rate in the second quarter. The speaker, David, views this as a positive indicator of competitive markets and the success of American companies.

Infrastructure and Energy Needs for AI

Scaling AI infrastructure presents significant challenges, particularly concerning power generation. The speaker emphasizes that working in the "world of atoms" (hardware and physical infrastructure) is inherently more complex than working in the "world of bits" (software). President Trump's "pro-energy" policies are credited with enabling increased power availability. Specific policy examples include:

• Support for the "drill, baby, drill" initiative, reflecting a farsighted understanding that energy is fundamental to all economic activity, including the AI boom.
• Allowing "behind the meter power generation," which enables companies to establish their own on-site power sources, reducing reliance on the main grid.
• Efforts to optimize the existing grid and promote new oil, gas, and nuclear power generation through an "all of the above" approach.

Government's Role in Private Sector Dealmaking

The White House maintains a policy of non-interference in specific dealmaking between private companies. David explicitly states that government officials, including himself, do not get involved in arrangements like the AMD and OpenAI collaboration. The government's role is to create a supportive policy environment that fosters competition and success for all American companies.

U.S. Strategy in the AI Race with China (The "China Hawk" Perspective)

David identifies himself as a "China hawk," asserting the critical importance of the U.S. winning the global AI race against China, which is considered the primary competitor. He argues that this race carries significant economic and national security ramifications. President Trump's strategy, outlined in a July 23rd speech, is based on four key "tentpoles":

1. Pro-innovation: Fostering technological advancement.
2. Pro-infrastructure: Building robust foundational systems.
3. Pro-energy: Ensuring sufficient power resources.
4. Pro-exports: Enabling the "American technology stack" to dominate the world. The overarching goal is to secure U.S. leadership and ensure its technology stack achieves global dominance.

Export Policy to China and Global Market Share

The U.S. export strategy should prioritize maximizing global market share for American technology. David criticizes the previous administration's policy of restricting chip sales to many countries, including resource-rich Gulf states, calling it a "major mistake." He argues that such restrictions compel these countries to turn to China and adopt the "Chinese tech stack," thereby ceding valuable market share. Conversely, President Trump's policy is seen as "boxing out China from the Middle East," whereas the previous approach "forced these countries into China's arms."

The primary metric for measuring success in the AI race is "global market share." The goal is for the American technology stack to achieve approximately 80% global market share within five years. A scenario where the Chinese technology stack (e.g., "Wylie chips and deep sea models") reaches 80% market share would signify a loss, akin to what "happened in 5G." Winning technology races, as understood in Silicon Valley, requires building the largest ecosystem, with the most users on the American technology stack possible.

The Debate on Exporting Deprecated Chips to China

The question of what to export to China is complex. While there is a strong argument against selling China the "latest and greatest chips" due to their potential benefit to China, a complete embargo could accelerate China's drive for technological independence. Therefore, a "compelling argument" exists for selling "deprecated American chips" – less advanced or older versions.

As an example, the Biden administration initially approved the export of Nvidia's "H-20" chips in "unlimited quantities" to China. When President Trump's administration took office, these H-20 sales were made subject to licensing and a 15% surcharge, a move that drew criticism despite the prior approval.

Talent Pipeline and Jensen Huang's Perspective

The U.S. must remain open to collaborating with global talent, including Chinese AI researchers, to some extent. It is noted that approximately "half the world's A.I. researchers are from China." Jensen Huang, CEO of Nvidia, emphasized the importance of this talent pipeline, a point David acknowledges as a nuanced question. Despite his "China hawk" stance, David was not "triggered" by Huang's comments, understanding them in their broader context. He defends Jensen Huang and Nvidia as a "huge American advantage" in the AI race, recognizing Huang's three decades of work on Graphics Processing Units (GPUs), which are crucial for AI computation.

---

Synthesis/Conclusion

The core of the discussion revolves around the United States' strategic imperative to win the global AI race against China. This involves a multi-faceted approach championed by the Trump administration, focusing on pro-innovation policies, robust energy infrastructure, and a strategic export policy designed to maximize the global market share of the "American technology stack." While advocating for technological dominance, the speaker acknowledges the complex balance required in export controls, suggesting that selling "deprecated" chips can prevent China from accelerating its self-sufficiency. Furthermore, the importance of global talent, including Chinese AI researchers, is recognized as a critical, albeit nuanced, factor in maintaining U.S. leadership, with a strong defense of key industry figures like Nvidia's Jensen Huang. The ultimate goal is to secure U.S. economic prosperity and national security through sustained technological leadership.