Kimi

NVIDIA's data center GPUs (Hopper and upcoming Blackwell) are the essential infrastructure for generative AI training and inference. The company has transformed into the picks-and-shovels provider for the AI revolution, with revenue increasingly driven by hyperscaler AI capex and enterprise adoption. Their full-stack approach (hardware + CUDA + networking via Mellanox) positions them as the foundational platform for AI workloads.

The CUDA software ecosystem represents one of the widest moats in technology, creating massive switching costs for developers and researchers. NVIDIA holds approximately 80-90% market share in AI accelerators with years of architectural leadership. The Mellanox acquisition added a networking moat for AI clusters, evolving NVIDIA from a chip vendor into a systems-level provider with platform lock-in.

["Valuation remains elevated and vulnerable to multiple compression if AI capex cycle slows","Geopolitical export restrictions on China sales create revenue headwinds and uncertainty","Hyperscalers developing custom silicon (Google TPU, Amazon Trainium, Microsoft Maia) could reduce long-term TAM","Potential cyclicality or inventory correction if AI infrastructure demand normalizes after initial buildout"]