For many enterprises, the use of artificial intelligence has evolved from an experimental phase to a necessary business expense, with digital transformation strategies expanding alongside AI innovation. However, many companies still struggle to fully grasp the unit economics of AI technologies, which can lead to financial risks as AI workflows scale and become integral to business operations. Cohere acknowledges these challenges and emphasizes the need for strategic decisions on whether to own or rent AI technology, noting that AI ownership is not a one-size-fits-all approach. Cohere said, "The total cost of ownership is the system that produces the token." This underscores the complexity of managing AI costs, which extend beyond visible token pricing to include hidden operational and infrastructure expenses.
The visible price of AI is typically quoted in tokens, which represent fragments of text processed by a model. Tokens are incurred across prompts, responses, and other AI interactions, yet token pricing is only part of the cost. The larger challenge lies in understanding the full cost of owning, operating, securing, and scaling AI once it becomes an essential business function. Cohere said, "AI ownership is not one size fits all." This highlights the need for enterprises to analyze which AI capabilities to own and which to rent, building the discipline to differentiate between the two.
According to a December 2025 IDC InfoBrief commissioned by DataRobot, 96% of organizations deploying generative AI and 92% using agentic AI faced higher-than-expected costs. These widespread overruns indicate that current planning models do not align with actual workloads. Gartner anticipates that AI will enter the "trough of disillusionment," where the main barrier to scaling is the unpredictability of returns, not lack of ambition.
Source: cohere
