$NVDA $MU $SNDK $LITE Another one bites the dust. Consolidation or simply turning off the lights is inevitable for many of these GAI "wrappers" . The frontier LLMs can natively do many of these functions very well, or customers are coding their own bespoke platforms. If I fast-forward 36 months, I feel like the frontier LLMs will need to capture as much margin in the service process as possible to effectively compete against the OSS LLMs. The majority of standalone "wrappers" will get squeezed and disintermediated out of existence. All the more reason to be invested in the GAI infrastructure trade. I believe the risk-adjusted ROIC is significantly higher in the GAI hardware stack than in the software/LLM layer. Yes, there will be one-off edge cases that invalidate this thesis.