Cerebras stock drops 50% as COO sells $78M after OpenAI shifts to Nvidia

Nvidia: Nvidia supplies leading GPUs that power the majority of AI training and inference applications. Its technology is now confirmed to support OpenAI's GPT-6 Astra Ultrafast tier, underscoring its role as the go-to expert for scalable, high-speed semiconductor solutions in AI. OpenAI: OpenAI creates frontier large language models and associated API services including premium speed tiers. It has tested specialized chip partnerships for inference optimization while selecting Nvidia hardware for delivering its newest Ultrafast model capabilities. Cerebras: Cerebras Systems develops specialized wafer-scale AI accelerators designed for high-performance inference workloads. The company had previously collaborated with OpenAI on fast inference tiers for earlier models but recent analysis shows OpenAI shifting the latest Ultrafast implementation to alternative hardware. GPT-6 Astra Ultrafast: GPT-6 Astra Ultrafast is OpenAI's high-speed service tier for an advanced GPT-6 model variant focused on rapid token generation. Nvidia has publicly stated that this tier runs on its GPUs, emphasizing performance gains in low-latency AI inference. AI Hardware Shift: OpenAI's choice of Nvidia GPUs for the latest Ultrafast tier follows prior use of Cerebras hardware on earlier model versions. Partnership Evolution: OpenAI continues to evaluate multiple hardware options as it scales inference infrastructure for speed-focused services. Semiconductor Expertise: Nvidia's GPUs have proven capable of supporting premium low-batch-size inference for frontier models previously associated with custom accelerators.