DeepSeek Claims to Threaten NVIDIA’s Moat—Here’s Why Jensen is Going to Keep Laughing to the Bank

Source: Nate’s Newsletter Date: 2025-01-27 URL: https://natesnewsletter.substack.com/p/deepseek-claims-to-threaten-nvidias

Summary

DeepSeek hitting #1 on the iOS App Store triggered sharp sell-offs in NVIDIA and Microsoft stock in late January 2025, driven by market assumptions that cheaper Chinese inference would reduce GPU demand. The author argues the sell-off reflects a category error — confusing efficient inference with reduced training compute demand — and that NVIDIA’s position is structurally intact because the Jevons paradox applies: lower per-token cost historically expands total consumption, not contracts it. The full technical argument is behind a paywall, but the thesis matches the broader industry read that followed.

Implications

• Feeds the compute economics / hardware thread: the DeepSeek moment crystallized the tension between efficiency gains and aggregate demand growth; this piece is an early articulation of the “demand expands to fill cheaper compute” counter-argument.
• Relevant to hardware investment theses and to understanding why NVIDIA guidance stayed strong through 2025 despite repeated efficiency-driven bear cases.
• Also a signal that public market participants consistently misprice efficiency improvements in AI — a recurring pattern worth tracking when similar announcements surface.