AMD DC GPU is only going to get interesting again in 26H2. MI400 and Rubin launch windows should be closer together (Both tied to HBM4 ramp). Although i'm guessing AMD will still be 1-2 quarters behind as tradition. Chiplet strategy only gets to be more superior as CoWoS goes to 5.5x reticle. ZT aquisition closed by then with rack scale solution ready. First uarch that is designed from the ground up with low precision / GEMM performance in mind. (if there is one thing AMD is alright at, it's uarch design). 1.5 more years to fix their shit software. UALink and their own inhouse AI optimized NICs should be ready by then.
MI355x is too little too late but I think MI400 is gonna give rubin a better fight than everyone currently expects..
2025 revenue & margins are currently severely overestimated by analysts but at the same time 2026 severely underestimated lol..
I agree. Although not sure if Nvidia ASIC (semi custom) group will do well. The big customers think Nvidia is too powerful. Good that Nvidia is trying but corporate politics might be a headwind.
Would be curious to hear your take on Alphawave Semi. They're not on many people's radar (being listed in the UK doesn't help) and given their positioning in the ASIC and networking space I'm surprised there hasn't been much chatter about them.
I think you mean MI355x at the end there?
AMD DC GPU is only going to get interesting again in 26H2. MI400 and Rubin launch windows should be closer together (Both tied to HBM4 ramp). Although i'm guessing AMD will still be 1-2 quarters behind as tradition. Chiplet strategy only gets to be more superior as CoWoS goes to 5.5x reticle. ZT aquisition closed by then with rack scale solution ready. First uarch that is designed from the ground up with low precision / GEMM performance in mind. (if there is one thing AMD is alright at, it's uarch design). 1.5 more years to fix their shit software. UALink and their own inhouse AI optimized NICs should be ready by then.
MI355x is too little too late but I think MI400 is gonna give rubin a better fight than everyone currently expects..
2025 revenue & margins are currently severely overestimated by analysts but at the same time 2026 severely underestimated lol..
Yea. AMD chose to mess with the naming. :(
Agree that MI355X is too late. We shall see how MI400X plays out.
I agree with you on amd problem for shorting it from here is that dc cpu and client seem to be doing really well again
I do believe that NVIDIA will find a way to resolve the issues eventually. The question is, when will that damn heat problem be fixed?
In the long term, NVIDIA is still very attractive. It has a strong moat and is even expanding into ASICs.
I have an analysis from a securities firm that I shared, and it mentions that NVIDIA might fall short of guidance until Q2. What do you think?
https://x.com/Jukanlosreve/status/1878300957093839343
I agree. Although not sure if Nvidia ASIC (semi custom) group will do well. The big customers think Nvidia is too powerful. Good that Nvidia is trying but corporate politics might be a headwind.
Would be curious to hear your take on Alphawave Semi. They're not on many people's radar (being listed in the UK doesn't help) and given their positioning in the ASIC and networking space I'm surprised there hasn't been much chatter about them.