Molecule_Invest

Semis is now firmly the FOMO trade. Regardless of fundamentals, every manager is feeling the pull of the seemingly infinite money glitch of owning semis. Lots of generalists I've spoken to in recent weeks have sharpened their pencils on the space, with some shifting meaningful portions of their book into select hardware names. So while no doubt pods are in semis and using internet/software as funding sources, the rotations of capitals are occurring in LOs that are obv very conscious of lagging benchmarks. Interestingly, the generalists venturing into semis are unconstrained by the knowledge/experience of how low quality some of these names were considered previously, or how cyclical the fundamentals of these stocks have been historically. Rather, it's easier to come to the space with a blank slate and form a view around the sustainability of the supply/demand imbalance in memory persisting. We have not done a great job at playing this. We own $NVDA and $ASML as they're quality businesses and we can get comfortable around the valuation. But they have much lower torque to the trade vs. say memory prices going parabolic and won't have the same insane earnings beats. We have been waiting for a pullback which very much has not occurred in some of the semi stocks we'd be comfortable owning.

Software largesse? 🧐🧐

Even Hock Tan admits it his custom ASIC customers HAVE to compete with NVIDIA. Every generation, Jensen raises the bar Everyone else is chasing Broadcom’s job? Help hyperscalers try to keep up That’s the strongest $NVDA bull case you’ll hear on a competitor’s earnings call

As we wrote in our TPUv7 article, TPUv7 will have competitive perf per TCO verus Blackwell in 2025 for large labs but TPUv8 will have worse perf per TCO verus Rubin in EOY 2026/2027. We added an subway surfer video as some people don't have the attention span to read an 10,000 word article. There will be 2 versions of TPU v8 but this will be different from the “P” (full) and “E” (lite) SKUs that featured in the then 4th and 5th generations. Rather, it is a dual track with one SKU co-designed with Broadcom (TPU 8AX ) and one SKU co-designed with MediaTek (TPU 8X). The strategy behind Google choosing to partner with MediaTek is to reduce the margin that they pay to their silicon design partners or specifically Broadcom. Broadcom charges Google for the whole System in Package which they stack a healthy margin on. This includes HBM. This is despite Google being largely responsible for both the front-end and back-end design of the compute elements of the chip, with Broadcom contributing various PHYs (most importantly Broadcom’s best in class SerDes) and controllers. Google wants to move to a model where their silicon partner only charges for the things they add value on and get closer to paying BOM cost for their silicon.