As of June 2024, Mizuho Securities estimates Nvidia controls 70-95% of the AI chip market for training and inference, supported by 78% gross margin and flagged CUDA software moat. However, hyperscalers (Google, Microsoft, Amazon) are deploying custom silicon for internal workloads, startups like Cerebras and D-Matrix are raising venture capital to build differentiated architectures, and AMD is gaining traction in inference-optimized applications with customer deployments at Microsoft. Structural shift underway from single-vendor dominance toward fragmented supply: device makers (Apple, Qualcomm) are also embedding neural processors for on-device AI, potentially shifting demand away from data-center GPU clusters. For financial institutions, this signals persistent Nvidia revenue concentration risk and material uncertainty around the addressable market for hyperscaler capex, since internal silicon deployment directly competes with external Nvidia GPU procurement.
579 other tracked developments touching the same claims.