Brian McCormick
10.4K posts

Brian McCormick
@bjmtweets
Exited Founder turned investor. Sharing business research. Fundamental asymmetric approach. Focus: $ROOT, $AMD, $KLAR, fintechs, insurtechs, and AI.















US NEW HOME SALES PLUNGED 17.6% MONTH OVER MONTH IN JANUARY TO 587K.



Are We Considering Ending Our All-In Position On $PATH? Join Aaron and Jake for a candid behind-the-scenes team meeting as we debate our latest portfolio positioning following recent earnings reports. We dive deep into the fundamental business models, future upside, and current market valuations of some of the most disruptive companies out there.


$LULU earnings day is always entertaining. Sentiment so washed and stock so beaten up. Makes you wonder how bad things need to be for stock to go meaningfully lower.





$AMD is actually on a favorable design path vs $NVDA. Nvidia pursued monolithic chip design instead of chiplets. This worked great as long as they could keep making the wafers bigger and TSM kept improving NM of chips. Yield may be worse this way, but efficiency and raw power is at its best. Which is why they are so attractive for training. However, TSM is at the theoretically physical limits of NM improvements (in fact, the smaller NM wafers are not even smaller now, it's just marketing where improvements are driven elsewhere). Nvidia's response is that since they can't make chips better with bigger or significantly improved wafers, they will start connecting duplicate dies together into one chip. The blackwell connects two duplicate dies together. They can continue this and connect 4 dies. But continuing on this trajectory to connect 8, 16, etc... is not a sustainable path. Whereas, $AMD is already a leader in chiplet design (connecting smaller purpose made chips together, not duplicates) and use of advanced packaging (adding height instead of width to a wafer) and using it to their advantage on memory. It also gives AMD better Total Cost of Ownership over Nvidia, and faster iteration speed of new chips, since they can use chiplet libraries to upgrade quicker over redesigning entirely new full wafers like Nvidia does for the most part. $AMD has already significantly narrowed the gap in their chips efficiency vs $NVDA, and is leading in multiple areas. Their biggest disadvantage is no true rack solution, which releases this summer, and will be used by OpenAi and Meta. New $AMD and $NVDA chips are now coming toward parity. With Nvidia's biggest lead being in interconnect speed. If $AMD can get a comparable solution, they could even take the lead... Or simply by iterating faster through chiplet release speed.













