Microsoft CEO Satya Nadella| Image Credit: BCCLFor years now, the technology industry has promoted artificial intelligence as a breakthrough that could dramatically improve productivity and reduce reliance on human labour. However, it is becoming increasingly clear to many businesses that adopting AI at scale is more financially challenging than initially thought.A recent report by Fortune has brought renewed attention, noting that the more workers use an AI-powered system, the more costly it becomes. This trend is already noticeable not only in startups but also in giants of the IT industry such as Microsoft and Uber, which seem to face rising costs associated with the use of artificial intelligence.Microsoft signals a wider industry concernAs reported in Fortune and The Verge, Microsoft has already begun cancelling most of the direct Claude Code licenses internally and is shifting its employees to the GitHub Copilot CLI solution. According to the news, several thousand Microsoft employees were initially encouraged to play around with Anthropic's coding assistant.The change reportedly comes just months after the company began offering more internal access to Claude Code. Although Microsoft did not disclose publicly that the main cause for this decision was cost, various sources claim that cost became an issue due to extensive use of the tool.Uber was also said to be spending more on artificial intelligence. According to reports, Uber's Chief Technology Officer, Praveen Neppalli Naga, informed employees that the company had already depleted its budget for 2026 AI code tools by the fourth month of the year. In addition to encouraging the increased adoption of AI coding technologies, the company used internal leaderboards measuring the use of such tools.Questions might arise whether the financial burden of utilising artificial intelligence will be a problem for firms looking to substitute or complement their workforce using AI in the long term.Why AI costs rise with usageUnlike traditional software subscriptions, most modern-day AI is based on token-based pricing. In other words, the more an organisation uses the services of AI systems, the greater the total expense will be.In its latest predictions for 2030, research company Gartner projects that performing inference on a trillion-parameter model would cost less than ten per cent of the same operation in 2025. In other words, as a result of increased efficiency of semiconductor technology, AI model architecture, and specialised hardware, individual AI transactions will become much cheaper over time.However, Gartner also warned that falling token prices may not reduce overall enterprise spending on AI. Reportedly, the next generation of what is termed “agentic” AI systems, which perform tasks with greater independence, need many more tokens to execute one job than ordinary chatbots do. Agentic AI systems could reportedly require 5 to 30 times more tokens per operation.Will Sommer, senior director analyst at Gartner, warned that companies “should not confuse the deflation of commodity tokens with the democratisation of frontier reasoning”.Simply put, despite the cost reduction of each individual AI task, the amount of such tasks done by enterprises could drive expenses through the roof.Image of Microsoft| Image Credit: Wikimedia CommonsResearchers warn of exploding token consumptionScholars have started exploring how AI autonomy will affect the economy. A recent research paper explored token consumption in agentic coding. The paper implied that AI agents performing sophisticated coding tasks could use upto 1000x more tokens than code chat or reasoning.However, researchers have established that more token consumption does not necessarily lead to better performance or increased accuracy. Accuracy would sometimes reach its peak point at moderate costs without improvement despite extra computing power expenditure.This is relevant since firms are increasingly developing AI systems designed to be somewhat independent within various processes, including customer support, coding, and enterprise work.Compute costs are becoming impossible to ignoreAccording to an interview, Bryan Catanzaro, Vice President of Applied Deep Learning at NVIDIA, stated that "the cost of compute is far beyond the costs of the employees." Thus, this statement contradicts one of the fundamental assumptions regarding generative AI: automation through replacement or support from AI will reduce operating expenses.On the contrary, the issue of increased costs related to productivity vs increasing expenses on cloud infrastructure, inference, and GPUs keeps emerging.Despite this, technology executives remain quite positive about the AI-powered future. Recently, NVIDIA CEO Jensen Huang said he believes every employee could eventually work alongside around 100 AI agents. However, if such a trend of increased usage of tokens persists despite lowering expenses, the situation may turn out differently.