Episode
How Chinas AI DeepSeek Is Upending the Model Cost Curve
- Published
- Jul 18, 2026
- Duration seconds
- 537
- Processing state
not_requested
Actions
POST https://stenobird.com/v1/public/podcasts/china-economy-with-fexingo-beijing-s-policies-trade-and-asia-pacific-markets-7871760/episodes/how-chinas-ai-deepseek-is-upending-the-model-cost-curve/transcription-requests
Idempotently request low-priority transcript generation for this episode.GET https://stenobird.com/podcast/china-economy-with-fexingo-beijing-s-policies-trade-and-asia-pacific-markets-7871760/how-chinas-ai-deepseek-is-upending-the-model-cost-curve.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
Episode 119 dives into how Chinese AI lab DeepSeek has quietly built a large language model costing around $5.6 million to train, versus the $100 million-plus price tags typical for frontier models from OpenAI and Google. Lucas and Luna discuss what this means for the cost structure of AI development, why DeepSeek's approach relies on architecture innovations from an overlooked research paper, and how the price gap could reshape the competitive dynamics between U.S. and Chinese tech companies — without needing trade sanctions or export controls. They also touch on the broader economic signal: if AI compute gets dramatically cheaper, it changes the investment thesis for the entire cloud sector. Market context includes a softening dollar and a yuan at 6.78, which makes Chinese exports even more competitive. The conversation stays grounded in a single concrete case and avoids sweeping predictions. #DeepSeek #AI #ChinaEconomy #ArtificialIntelligence #LanguageModel #CostEfficiency #TechCompetition #OpenSourceAI #CloudComputing #NVIDIA #MoEArchitecture #ModelTraining #ExportControls #Yuan #TradeCompetitiveness #FexingoBusiness #BusinessPodcast #Economics Keep every episode free: buymeacoffee.com/fexingo