Six-Chart Sunday (#54) – Bad AIssumptions
6 Infographics + 1 Video (Sahil Bloom's "The 5 Types of Wealth")
The Great DeepSeek Freakout of 2025 is upon us. (In case you’ve been living under a rock, a Chinese hedge fund released an impressive new generative AI model built & operated at a fraction of the cost of leading U.S. systems and despite semiconductor export controls designed to slow down China’s AI progress). Markets convulsed, Cold Warriors flexed and pundits declared a new “Sputnik Moment.” This seems overwrought & historically-off. The more one reads about DeepSeek, the more it seems a positive signpost for the AI future, upending wrongheaded assumptions.
Assumption: Sputnik moment. Reality: Toyota moment. The Soviet launch of Sputnik in 1957 showed the U.S. had (a) fallen behind (b) an enemy government on (c) fundamentally military technology with (d) a secretive system whose capabilities we did not understand. By contrast DeepSeek is a fully-transparent (open source), non-military offering developed by a non-governmental player that trails U.S. leaders (from whom they borrowed). A better analogy may be the Toyota Moment… when America’s auto industry realized upstart Japanese competitors could make more fuel efficient & reliable cars at lower cost, forcing serious rethinking of U.S. investment, workforce, processes, trade and market strategies for a critical sector.
Assumption: Big Tech has Deep Moats. Reality: Big Tech has Narrow Moats. The Magnificent Seven AI stocks (led by NVIDIA) have been unstoppable as we assumed bigger AI was always better. NVIDIA’s stock suffered the biggest one-day loss of market cap in world history when markets grasped DeepSeek’s capabilities despite lesser hardware & investment, while open source AI leaders such as IBM soared.
Assumption: Game-changing risk revealed. Reality: Game-changing benefits unlocked. “For AI to transform society, it needs to be cheap, ubiquitous and out of the control of any one country or company. DeepSeek’s success suggests that such a world is imaginable.” (The Economist). Effective lower-cost models will diffuse faster and make higher productivity outcomes more likely for the U.S. and world.
Assumption: China trails in innovation and always will (due to lack of freedom). Reality: China leads in many areas and often has. China graduates more science, technology, engineering, and mathematics students than the rest of the world combined, accounts for nearly half of the world’s top-tier AI researchers and has pulled ahead in 5G, high-speed rail, electric vehicles, batteries, solar panels and drones, among other tech.
Assumption: More efficient AI will ease surging electricity demand. Reality: More efficient AI may increase energy demand & LLM usage. While some observers suggest “DeepSeek’s more efficient model threatens energy demand growth forecasts,” others see this as a classic Jevons Paradox… if AI is more efficient, it will be used more, so energy demand will still grow.
Assumption: We must control more. Reality: We must run faster. DeepSeek may prove a wake-up call for Washington policymakers concerned with American economic, technological, military & global leadership. While some are calling for tighter controls on technology, scientists & spending, close observers of the first Cold War understand the West won by running faster… innovating faster (via serious government investments in R&D), attracting & retaining more elite global talent (via STEM education & skilled immigration), leveraging the whole economy, not just the defense sector (via proliferation of open source & dual use technologies) and aligning with allies (via constructive trade & defense partnerships).
Upcoming live discussion on tax legislation, Friday 2/7 at 2pm ET, register HERE
VIDEO
As Arthur Brooks noted, "You've always heard that money doesn't buy happiness, but you almost never hear what you should accumulate instead. Sahil Bloom masterfully answers this question." Sahil & I discussed his great new book this week:
I am a big fan of Bruce's newsletter. Always enlightening and informative. Regarding how the U.S. won the First Cold War, there are two comments: 1. By naming "The First Cold War", we assume there will be a Second, if not a Third. I would challenge such an assumption, although I realize this has become a consensus in the U.S. I hope we can turn this mindset around before it evolves into a reality, which would be unfortunate for the world (including the U.S.) because China is not the Soviet Union. 2. Instead of concluding the U.S. won the First Cold War, it is equally, if not more, appropriate to argue that the Soviets lost. The U.S. was lucky to have the Soviets as an adversary.
Such a timely, well thought-out assessment. We need more unconventional wisdom like this.