Why the US can't ignore the China AI threat

Key Concepts

• Hyperscalers: Large-scale cloud service providers (Microsoft, Amazon, Google).
• Closed-Source AI Models: AI models developed and controlled by private companies (OpenAI, Google, Anthropic, XAI).
• Open-Source AI Models: AI models publicly available, allowing for community contribution and modification, prevalent in China.
• Blackwell & Reuben: Nvidia’s latest generation of GPUs (Graphics Processing Units) used for AI processing.
• AI Bid: Investor enthusiasm and increased stock valuation driven by perceived AI potential.
• Infrastructure (Dirty Infrastructure): The physical components supporting AI, including data centers and power generation, with "dirty infrastructure" referring to power sources with environmental concerns.

---

Political Pushback on Data Centers & Microsoft’s Response

The discussion began with the increasing political scrutiny surrounding data centers, highlighted by a recent article noting the political left’s concerns over high power prices, particularly voiced by Bernie Sanders. This pressure prompted Microsoft to proactively address the issue, recognizing the tension between data center expansion and political concerns. Doug believes this tension will persist for years, but the “AI train is just so hard to stop,” suggesting limited political willingness to hinder US leadership in AI. He emphasized that hyperscalers like Microsoft, Amazon, and Google have experience navigating community concerns during data center builds.

China’s AI Advancement & Competitive Landscape

Microsoft President Brad Smith’s commentary on China’s AI progress was a central topic. Smith stated that “China is winning the AI race outside of the West,” due to its competitive open-source models and government subsidies allowing them to undercut American companies on price. Doug affirmed Smith’s assessment, framing the AI race as a “tale of two players.”

He differentiated between the US’s lead in closed-source AI models (OpenAI, Google, Anthropic, XAI) and China’s strength in open-source models. A key advantage for China, according to Doug, is its greater willingness to build infrastructure, including “dirty infrastructure” – prioritizing data center power regardless of the environmental impact – and its investment in nuclear power, unlike the US which hasn’t built a new nuclear plant in years. This infrastructure advantage is crucial for supporting AI development.

Nvidia Chip Access & China’s Regulations

The discussion then turned to a report detailing the Chinese government’s restrictions on tech companies purchasing Nvidia’s H200 AI chips, limiting approvals to cases like university research. Doug anticipates ongoing negotiations, but believes Nvidia chips will eventually return to China. He views this as a potential “tailwind” for Nvidia, as the possibility of Chinese access isn’t currently factored into the stock price. He suggests China will structure regulations to appear favorable to its constituents while still securing necessary chips for open-source model development and competition with the US.

Nvidia’s Stock Performance & Investor Allocation

The conversation addressed Nvidia’s stock price, which has been trading in a relatively narrow range for six months. Doug attributes this to a lack of new catalysts, despite positive fundamentals like Jensen Huang’s announcement of a $500 billion backlog for Blackwell and Reuben GPUs (continuing to grow as of CES). He believes the primary issue is that many investors are already “fully allocated” to Nvidia, limiting the potential for further price movement from a mega-cap perspective.

Intelligent Alpha’s Portfolio Strategy: Amazon vs. Nvidia

Doug revealed that Intelligent Alpha, his firm, favors Amazon over Nvidia currently. He drew a parallel to Google’s situation in mid-2023, where negative sentiment was fully priced in, and a new model release (Gemini 2.5) triggered a stock turnaround. He believes Amazon is in a similar position, having been the worst-performing mega-cap last year (up only 5%). He argues that there’s “no AI bid” in Amazon’s stock yet, but strong AWS performance could attract AI-focused investors and drive the price higher. He specifically noted that their GPT models favored Amazon.

Logical Connections

The discussion flowed logically from the initial political pressures on data centers to the broader geopolitical context of the US-China AI race. The conversation then narrowed to specific companies (Nvidia, Microsoft, Amazon) and their respective positions within this landscape. The analysis of Nvidia’s stock performance and Intelligent Alpha’s investment strategy provided a practical application of the broader themes discussed.

Data & Statistics

• Nvidia Backlog: $500 billion (for Blackwell and Reuben GPUs).
• Amazon Stock Performance (2023): Up 5%, the worst-performing mega-cap.
• Gemini 2.5: Google’s model release that spurred a stock turnaround.

Conclusion

The conversation highlighted the complex interplay between political pressures, technological advancements, and geopolitical competition in the AI landscape. While the US currently leads in closed-source AI, China is rapidly gaining ground in open-source AI, fueled by government support and a willingness to prioritize infrastructure development. Nvidia remains a key player, but its stock performance is constrained by investor allocation. Intelligent Alpha’s preference for Amazon reflects a belief that the market hasn’t fully recognized the company’s AI potential, mirroring a previous turnaround seen with Google. The overall takeaway is that the AI race is far from over, and navigating this evolving landscape requires a nuanced understanding of both technological and political factors.