Full HTML article body
The pace of AI infrastructure investment just accelerated again. Meta is preparing to begin manufacturing its own AI chips by September, targeting 14 gigawatts of compute. Meanwhile, OpenAI, xAI, and others are reshaping agentic AI and cloud infrastructure.
But the same time, OpenAI and xAI are rolling out agentic and voice models at a rapid clip, signaling that the AI industry is moving simultaneously on two fronts: raw compute and application-layer infrastructure.
For cloud computing observers, the implication is clear: the hyper-scalers are increasingly designing their own chips, and Amazon (Trainium/Inferentia) is leading the charge. This scale is not just about the largest private AI compute farms, but also about the larger private infrastructure that supports the application layer.
For cloud computing observers, the implication is clear: the hyper-scalers are increasingly designing their own chips, and Amazon (Trainium/Inferentia) is leading the charge. This scale is not just about the largest private AI compute farms, but also about the larger private infrastructure that supports the application layer.
For cloud computing observers, the implication is clear: the hyper-scalers are increasingly designing their own chips, and Amazon (Trainium/Inferentia) is leading the charge. This scale is not just about the largest private AI compute farms, but also about the larger private infrastructure that supports the application layer.