The AI race looks like it might be moving into a new phase.
Lisa Su mentioned that 2026 could be the first year where more AI compute is used to run models than to train them. That shift actually matters.
As AI agents get adopted more widely, inference demand could turn into the next major growth driver. She expects inference to consume around 60% of global AI capacity.
The market has been so focused on training chips, but the next opportunity might come from the massive computing power needed after models are built.
For $Advanced Micro Devices(AMD)$ , the real question is execution — can they grab more AI accelerator share and stay competitive in the next stage of the AI infrastructure cycle? I'm watching this one closely.
AI isn't slowing down, it's evolving. $Advanced Micro Devices(AMD)$ is on my radar.
Comments