THE DEEPSEEK MOMENT....AND TARIFFS
Hello and thanks for joining me for this week’s edition of the No Straight Lines Investments blog, I daresay this week’s market action validated my choice of names for the blog!
By now we are all aware that NVDA shedding $650B in market cap in a day is a historical record.
NVDA is the undisputed champion of the AI trade thus far, so it is notable that the stock finished the week down 15.8%, even after other names levered to the so-called infrastructure build out phase of AI bounced back (more in the Charts of the Week below).
Notwithstanding debates regarding the “true” cost of training and inferencing for DeepSeek’s R1, what are the conclusions, thus far, about the DeepSeek effect:
DeepSeek’s use of reinforced learning, mixture-of-experts/multi-head latent attention (fewer parameters) and distillation from other robust LLM’s is a roadmap that every AI player is now utilizing.
As Mark Lipacis of EvercoreISI points out, technology history shows that lower cost/compute cycles = more demand for compute cycles.
But you don’t have to take Mr. Lipacis’ word (or this humble blogger’s), it is already showing up in higher rental costs for H100 and H200 compute (from Dylan Patel, SemiAnalysis).
If you would like to continue reading, please subscribe for $20/month using the button:
Here’s what’s in store:
Portfolio Stocks - focus on software, uranium, China and hardware. My hope is to introduce readers to ideas that are not necessarily on the front page, yet represent fantastic risk/reward profiles.
Charts of the Week - post DeepSeek, what conclusions could be drawn from price action, Q4 earnings winners and losers and more!
Flows - where are retail and professional investors allocating their investment .Ifyoudon′tfollowthe you are flying somewhat blindly.
Macro that Matters - took a bit of a backseat, but between BofC and FOMC and tariffs, lots to discuss.
If this is it for you, thanks for reading and best of luck with your investments. You can follow me on X (@NSLInvestments), LinkedIn: Jonathan Lansky, or using the chat feature on the Substack app.