Fort Washington Investment Advisors · 2025 · Published 6 February 2025
Christopher Shipley, Sunit Gogia and Blake Stanislaw argue that DeepSeek's cheaper open-source model makes Nvidia's chip demand less predictable in the short run but should widen AI adoption and raise long-run chip demand, as Jevons paradox predicts.
The February 2025 Market Pulse, by Christopher D. Shipley, Sunit Gogia and Blake W. Stanislaw, covers a January in which the S&P 500 gained 3% even as DeepSeek's model hit U.S. semiconductor shares led by Nvidia. Value beat growth by over 3% for the month. The authors expect the 25% tariffs announced on Mexico and Canada to be political and temporary, and tariffs on China to matter more over time. The Fed held rates after cutting 100 basis points from September to December, while the 10-year yield sits nearly 1% above its level before the first cut. The Spotlight argues that falling AI training costs will spread adoption and that scaling laws still hold.
While there are still some outstanding questions around DeepSeek’s release, the company’s open-sourced model showed meaningful improvements in cost efficiency compared to U.S.-based models (OpenAI, Meta, etc.).Show in the letter ›
We anticipate that as LLMs (Large Language Models, such as Open AI’s GPT-4 or Meta’s Llama) and other AI models become widely available at low costs, companies will leverage them to build custom solutions for their specific business needs, which will drive productivity gains.Show in the letter ›