Okay, 3 out of the 4 @BanklessHQ podcasts this week talked about DeepSeek's R1 model ======== After all the interviews, here are my summarized thoughts: 1) At worst, the introduction of efficient inference models like R1 is a small L for Nvidia, because inference costs will approach 0 faster than prev expected. 2) At best, it's actually a solid W for Nvidia, because cheap inference will induce demand for more inference ("Jevons Paradox") 3) Open-source AI takes the biggest W, since it shows that proprietary models will quickly be copied, and ultimately become open source commodities 4) Consumers and the economy also take a very large W, because - The AI Arms race increases in speed & funding - All AI products become cheaper, quicker, and more applicable 5) The NVDA pullback is normal and expected, compared to the size of its run up. Nvidia is getting unbundled from multiple directions. DeepSeek was a coincidence that sparked a fire. It wasn't the market realizing it was over-indexing on compute, the market realized it over-indexed on NVDA's monopoly more broadly. 6) Apple takes a W, because R1 makes locally-run AI inference more feasible. 7) The accelerated commoditization of models shows that proprietary data is the most valuable edge that exists in the AI market - Data is where the line will be drawn between the U.S. and China. Compute is a fungible commodity, but having data that your competitor doesn't is priceless.