OpenAI confirms 750 MW of Cerebras inference capacity through 2028 multi-tranche
Following the May 14 Cerebras IPO, OpenAI provided unusual detail on its deployment plans: 750 megawatts of Cerebras-based inference capacity will come online across multiple tranches through 2028, with the first 100 MW already in production at Cerebras's Memphis site.
The disclosure adds concrete numbers to the previously-announced "$20B over multiple years" deal. 750 MW translates to substantial real-world inference throughput — likely in the range of tens of billions of tokens per second across the deployment.
The multi-tranche structure protects both sides: OpenAI gets supply diversification away from NVIDIA, Cerebras gets predictable demand that justifies the wafer-scale fab capacity buildout.