The AI subsidy is ending, and waste is about to become a line item (www.speakeasy.com)

The era of subsidized AI is coming to an end, significantly impacting enterprise budgeting for AI/ML applications. Both Anthropic and OpenAI have filed for IPOs, revealing a shift in pricing strategies where products are now priced for profitability rather than widespread adoption. For example, OpenAI’s latest model costs double that of its predecessor per token, reflecting a broader trend where major AI coding vendors are moving away from all-you-can-eat pricing models to usage-based billing. This transition has already led to unexpected budget constraints for enterprises that had projected their 2026 budgets based on previous pricing models. The implications of this shift are profound for the AI/ML community. Organizations are now facing rising costs associated with "AI waste," where inefficiencies in token usage can inflate bills. Factors such as increased reliance on frontier models, a surge in inference workloads, and the transition from flat-rate to metered pricing are driving costs higher. As companies look to maintain their AI operations amidst these rising expenses, the focus is shifting toward implementing an AI Control Plane to improve cost management. This infrastructure—capable of monitoring and optimizing token usage—will be vital for organizations aiming to manage AI spend effectively, ensuring that every dollar allocated to AI directly correlates with productive output while mitigating waste.