Quote · The Prof G Pod with Scott Galloway
China Decode: Why China Got Locked Out of SpaceX and America’s Biggest IPOs (ft. Ed Elson)
Where this was said
US Data Center CapEx vs China's Lower-Cost Buildout — And the Numbers That Mislead
At 1:10:12 · chapter starts 1:08:10
Alice Han challenges the common narrative that the US is comfortably ahead of China on AI infrastructure investment. US data center CapEx is projected at $750 billion in 2026 alone, while China's announced target is roughly half that by 2030 — numbers that seem to show a clear US lead. But Alice argues this comparison is deeply misleading because infrastructure and energy buildout costs per gigawatt are dramatically lower in China, meaning the actual capacity gap is far smaller than headline spending figures imply. Ed Elson extends the point to the model layer, noting that Deepseek's AI models are reportedly 96% cheaper than OpenAI equivalents — and that as US enterprises face spending caps on AI, cheaper Chinese alternatives become increasingly attractive. Alice delivers a striking anecdote: major US tech companies she encountered at a Hollywood conference are actively using the Chinese open-source model Qwen because it's cheaper, faster, and easier to fine-tune than OpenAI or Claude. The great AI decoupling, it turns out, is not happening at the enterprise level.
The US-China AI decoupling narrative has a problem: it isn't happening at the enterprise level. Alice Han reports that major US tech companies she encountered are actively using Chinese open-source models like Qwen because they're dramatically cheaper than OpenAI or Claude. Deepseek's models are reportedly 96% cheaper than OpenAI's equivalent.
Deepseek's AI models are reportedly 96% cheaper than comparable OpenAI models, driving growing interest in Chinese AI alternatives among cost-conscious enterprises.
The US is projected to spend at least $750 billion on data center infrastructure in 2026, roughly twice China's stated target of approximately $375 billion by 2030.
