Cerebras Systems CEO Andrew Feldman announced that OpenAI’s GPT-5.4 model is now running on Cerebras hardware, with plans for GPT-5.5 rollout. This partnership, formalized on January 14, 2026, under a multibillion-dollar agreement exceeding $10 billion, will ensure 750 MW of compute capacity through 2029.
The first milestone was the February 12, 2026, launch of the GPT-5.3-Codex-Spark model, achieving over 1,000 tokens per second. With GPT-5.4 confirmed live and GPT-5.5 in development, Cerebras is positioning itself as the preferred inference engine for OpenAI’s next-generation products.
Through its innovative wafer-scale engine, Cerebras enhances AI interactions, reducing communication overhead typical in multi-chip setups. The company completed its IPO in May 2026, valued between $63 billion and $95 billion, assisted significantly by the OpenAI partnership.
Public release of GPT-5.4 and GPT-5.5 is expected shortly, backed by the substantial compute capacity deal. However, the concentration risk of depending on a single customer, OpenAI, is a critical factor to watch.