U.S. Options to Outcompete China
NOTE: The opinions expressed here belong solely to me and may not reflect the views of my employer.
A few days ago, China announced a $137 billion (1 trillion yuan) investment over five years, to compete against the United States "Stargate Project."
China's plan is fully state-backed, with supporting firms like Baidu, ByteDance, Alibaba, etc. China announced a new AI company, “DeepSeek”. Today, the stock market faced a historic disruption as $2 trillion in value evaporated. Now, the United States faces a pivotal moment in deciding how to respond.
Here’s my rough analysis of potential strategies:
Option A: Formal “Declaration of AI Race” by U.S Government
In moments of great ambiguity, nations thrive or falter based on their ability to rally around a shared vision. For the United States, the concept of a unifying vision isn’t just about survival or progress — it’s about finding a common purpose that transcends divisions and compels us to explore, and confront, the fundamental questions of existence. A shared vision has the power to unite a country, not merely through strategy, but through a collective understanding of who we are, why we strive, and what it means to be human in an age of rapid transformation.
If a unifying vision is needed, President Trump could follow historical precedent, by requesting a technology race, through Congress. This would involve committing to a bold, strategic investment in AI development, similar to President John F. Kennedy’s historic address to Congress on May 25, 1961. In his speech on “urgent national needs,” JFK called for investment in the space program, proclaiming that “this nation should commit itself to achieving the goal” of landing a man on the moon within the decade. A similar declaration today, could empower the United States.
If this a formal race was fully funded by taxpayer dollars (like China), while carrying the potential to disrupt American jobs via advanced AI-automation… it seems unlikely to be appealing among the American public. However, if continued to be funded by private capital and shared corporate interests (ex: project stargate), including all (Meta, Apple, Oracle, IBM, OpenAI, Tesla, Google, Anthropic, AWS, etc… it would be palatable.
Pros: This approach could mitigate some of the current economic challenges, such as stagnant innovation cycles within the United States Government, and lagging onshore job growth in emerging tech sectors. By fostering AI investments across all major AI companies, innovation can be accelerated.
Cons: However, such a move risks exacerbating inflationary pressures in some sectors, or frightening the American people, while also heightening tensions with China, potentially triggering broader geopolitical ramifications.
Option B: Establish a Manhattan-Style AI Mission
The United States could take a nuanced inspiration from the Manhattan Project, a pivotal initiative that changed the course of history during World War II. The Manhattan Project was born out of the existing boom in radiological research and a growing fear that another nation could weaponize this technology first. Recognizing the stakes, the United States government launched a highly coordinated and secretive effort to recruit the brightest minds from academia and across the globe, consolidating their expertise into a singular, goal-driven mission.
A modern parallel can be drawn to the current boom in AI research within American academia and private industry. Today, academic institutions across the country are focusing heavily on various domains of Artificial intelligence research. Likewise, leading tech companies compete fiercely to attract top AI talent, resulting in siloed innovation rather than collective progress. A Manhattan-style project for AI would unite academia and various AI companies under a shared initiative, pooling public and private resources, expertise, and incentive, to achieve a monumental goal.
By stoking collaboration among AI stakeholders with distributed governance efforts, and forging continent-wide clusters, into a unified mission, this strategy has the potential to unlock unprecedented innovation while minimizing cost to taxpayers. Tactically speaking, this could take the form of a small United States government agency being embedded within specific AI companies to oversee and lead “the mission”, centrally coordinating an exchange of advancements across teams. Alternatively, a more pragmatic and scalable approach for collaboration, could look like incentivizing individuals within academia and AI companies to share relevant research, insights, and advancements, via a new/special clearance issued from the United States.
Pros: Such a project would break down competitive silos, enabling unprecedented advancements in AI, Artificial General Intelligence (AGI), and Super intelligent AI (SAI). By uniting the brightest minds and leveraging existing momentum in AI research, the United States could accelerate AGI / SAI. It would also remain outside the public eye, therefore reduce domestic fears, and conceal market shaking geopolitical tensions.
Cons: The potential risks of such a large-scale initiative remain uncertain. The ethical considerations, and potential geopolitical ramifications of consolidating AI efforts into a single project would require careful navigation to avoid unintended consequences.
Option C: Price Controls and Cost Benchmarks
Another potential path involves Congress passing legislation to mandate price controls for GPUs, or, cost-efficiency benchmarks for AI models developed in the United States private sector, while simultaneously pursuing greater raw material excavation. This approach could lower the cost of AI for consumers and enterprise organizations who have yet to adopt the technology, enabling broader proliferation.
GPUs, the backbone of AI development, can be likened to gasoline — essential and scarce, during times of crisis. In an AI emergency, where GPUs become critically scarce, there is historical precedent for price controls. For example, the Consumer Fuel Price Gouging Prevention Act, passed by the House of Representatives in May 2022, which empowers the President to declare an energy emergency and restrict “unconscionably excessive” pricing. Similarly, Congress could pass legislation to limit GPU prices during an AI crisis, ensuring affordability. To further address the issue, lawmakers could simultaneously pass policies to unlock raw materials via increasing rare mineral excavation, such as lithium and cobalt, to reduce the cost of GPU production.
However, it is important to recognize the inherent risks of imposing price controls or cost benchmarks on such a nascent and dynamic sector. Historically, these have proven to be dangerous to economies, often stifling innovation and creating inefficiencies. Reducing profit margins for semiconductor companies, or AI software companies, could significantly slow innovation, as the incentive to invest diminishes. This is particularly concerning in AI software companies, where margins are already razor-thin due to the extraordinary costs of GPU infrastructure and training large models. Unlike standard software products, which often yield high margins, semiconductors and AI model development are complex, making profitability elusive, even for established companies.
Pros: This approach could make AI more affordable and accessible to end users in the short-term, maybe.
Cons: Mandating cost controls risks de-incentivizing private investment in AI, reducing startups' ability to compete, and discouraging further innovation. The already thin profit margins in AI, compounded by high GPU costs, mean that companies may deprioritize cutting-edge research.
Option D: Leverage Deflationary Benefits of Loss
Alternatively, the United States could take a counterintuitive approach: thank China for inadvertently reducing the global supply of dollars, thereby easing inflationary pressures domestically. The United States could then pull back on federally funded AI initiatives to double the effect.
Pros: This strategy emphasizes de-escalation, cooling the AI race, and acknowledges an unintended positive economic outcome.
Cons: No significant drawbacks arise from this approach, apart from possibly appearing weak/incompetent.
Option E: Encourage Escalation Measures
The United States could encourage escalation, with a gentle nod toward the tech sector, to further intimidate China using uncoordinated style.
Pros: This approach would send a message, signaling United States control of technological persistence and resolve.
Cons: However, such actions would not only upset China, but also the broader international community.
Option F: Focus on Cost Optimization
The United States could take no action at all, and allow AI companies to cost optimize at their own pace and make models more efficient, which is already underway.
Pros: This approach would suggest the stock market has nothing to worry about.
Cons: The downside of this option is unknown.
Option G: Focus on Open-Source
The United States could encourage AI companies to invest more resources into Open-Source AI models.
Pros: This approach would expand access to AI technologies and allow broader competition, enabling community development, better security, interoperability, flexibility, and longevity.
Cons: Open-source AI could reduce the value of the managed services they compete against, in the short-term.
Conclusion
BigTech is already responding with option (F), so I personally like the idea of option (B)… and maybe (G).
The Manhattan approach is particularly attractive because it addresses a key barrier to innovation: distributed governance incentives with a shared vision across the academic and private sector. By introducing mission-centric governance, or, emboldening a tantalizing and highly scalable bottom-up collaboration, the United States can create the conditions for rapid progress in AI. Unified effort could accelerate advancements in areas such as Artificial General Intelligence (AGI) and Super intelligent AI (SAI), ensuring that the country remains a global leader in technological innovation.
The concept of a decentralized "Manhattan Project" for AI develop is intriguing, precisely because of its complexity and scale. According to a CSET analysis of ACS data, 10.4% of the United States workforce was employed in AI-related occupations in 2022 — at the height of tech employment, before layoffs began. Even conservatively assuming that 5% of the workforce remains tied to AI and that 1% of those are "top scientists" (a generous assumption), such an initiative would need to oversee approximately 60,000 lead scientists and potentially three times that number in supporting staff. This suggests a total of around 180,000 individuals who would need to be engaged and governed — a significant undertaking by any measure. Source
If it were up to me, here’s my high-level blueprint to get the conversation started.
The current state of AI innovation is at a pivotal juncture, do we press forward or pause? There is ample time for this hot-topic to be cooled. As the government evaluates its options, it must weigh the benefits of “inflating and pursuing” a new race, against the option of “deflating and diverting” attention. If the objective is to drive AI development further, then a coordinated, collaborative effort akin to the Manhattan Project offers a beginners framework.
It’s difficult to imagine the United States veering away from a future shaped by AGI (Artificial General Intelligence) or SAI (Superintelligent AI). This trajectory feels inevitable, given that our infrastructure is designed — and continues to evolve — to prioritize superior innovation. Fun fact, the United States hosts “5,381 data centers” within its borders, a staggering “10x” the number located in China. Source
