Cerebras Systems Inc. Class A Common StockReport that OpenAI's Ultrafast tier uses Nvidia GPUs rather than Cerebras chips challenges Cerebras' architectural advantage and growth narrative.
Cerebras Systems shares tumbled 7% after semiconductor research firm SemiAnalysis posted on X that OpenAI's newly announced GPT-6.1 Sol Ultrafast tier is running on traditional Nvidia GPUs at a low batch size rather than on Cerebras chips. OpenAI's Ultrafast tier promises token generation speeds up to eight times faster than standard deployments, reaching an estimated 300 tokens per second, a workload requiring the massive memory bandwidth Cerebras built its Wafer-Scale Engine to deliver. The report's focus on low batch size directly challenges Cerebras' architectural advantage, since Nvidia GPUs typically need large batch sizes for peak efficiency while low batch size prioritizes immediate response speed. Analysts immediately questioned whether Cerebras will secure any role in serving the GPT-6.1 Sol Ultrafast model in the future, a blow to near-term investor confidence in the company's growth narrative. Cerebras has yet to issue a public statement confirming or denying its involvement in OpenAI's future infrastructure roadmap.
Cerebras Systems Inc. Class A Common StockReport that OpenAI's Ultrafast tier uses Nvidia GPUs rather than Cerebras chips challenges Cerebras' architectural advantage and growth narrative.
NVIDIA CorporationReport says OpenAI's GPT-6.1 Sol Ultrafast tier runs on traditional Nvidia GPUs, a competitive win over Cerebras for this workload.
OpenAI is the subject whose GPT-6.1 Sol Ultrafast tier is being served on Nvidia GPUs, but the article reports no direct positive or negative development for OpenAI itself.
SemiAnalysis is only the source of the report on X, not a company affected by the news.