OpenAI's New Codex Version Powered by Dedicated Cerebras Chip

OpenAI is diversifying its technological foundations in a significant hardware shift. A new version of its Codex system is now powered by a dedicated, custom chip from Cerebras Systems, marking the 'first milestone' in a strategic partnership between the two companies. This move represents OpenAI's first major inference deployment that steps beyond its traditional, almost exclusive, reliance on NVIDIA's GPU infrastructure. The collaboration with Cerebras, known for its wafer-scale engine chips, is aimed at achieving greater efficiency and performance for specific, demanding workloads—in this case, AI-powered code generation. Using purpose-built hardware can offer advantages in speed and cost-per-inference compared to general-purpose GPUs. This strategic diversification gives OpenAI more control over its stack, potentially reduces costs, and mitigates supply chain risks. It also signals a growing trend in the AI industry: as models and their applications become more specialized, so to