Nvidia's New Hedge Against Chip Competitors? Partner with Them

OpenAI has unveiled a custom inference chip called Jalapeño developed with Broadcom, joining Google, Apple and SpaceX in moving away from reliance on Nvidia hardware, according to TechCrunch reporting. The announcement represents a declaration of independence from the GPU maker that has controlled AI infrastructure pricing and availability for years, the report said. While Nvidia still commands over 80% of the AI accelerator market according to industry analysts, that grip is loosening as major players bet billions that building their own silicon beats waiting for GPUs. Companies running massive AI inference workloads face expensive and backordered Nvidia chips designed for training rather than serving billions of daily requests, the report said. A chip optimized purely for inference can deliver the same performance at a fraction of the cost and power consumption. Broadcom has become the go-to partner for companies wanting custom silicon without building their own fabrication plants. Google used Broadcom for recent TPU generations and OpenAI is following the same playbook. Nvidia recently announced its Blackwell architecture with inference-optimized features and is slashing prices on older generations to maintain market share. CEO Jensen Huang has publicly dismissed custom chip efforts as expensive distractions that cannot match Nvidia's economies of scale and software ecosystem. The strategic calculation is clear: AI companies building trillion-dollar valuations cannot afford to have infrastructure costs controlled by a single supplier.