- Businesses that once aggressively pushed employees to use AI are starting to tighten usage due to rising costs.
- Amazon, Walmart, Cisco, Uber, and Meta are among the companies that have implemented AI spending controls.
- Some enterprises are applying usage limits, encouraging the use of cheaper models, or preventing wasteful AI usage.
- This trend marks a new phase in enterprise AI deployment.
- Previously, the vast majority of employees mainly used AI chatbots.
- Currently, an increasing number of companies are deploying AI Agents capable of automatically executing complex tasks.
- However, AI Agents consume significantly larger amounts of computing resources compared to conventional chatbots.
- Cost pressures rose sharply after OpenAI and Anthropic shifted many services from a fixed-subscription model to token-based pricing.
- Deloitte expert Costi Perricos stated that AI computing costs have now become a matter of concern for CFOs and boards of directors.
- OpenAI CEO Sam Altman also admitted that cost has become “a very big issue” for customers in 2026.
- Uber stated it is increasingly difficult to prove that AI spending is generating proportionate value. The company exhausted its entire 2026 AI budget by April. Consequently, Uber applied a limit of around 1,500 USD in token costs per month for each AI tool used by employees.
- Walmart is also limiting the number of tokens used on its internal AI systems. Walmart’s CTO noted that the usage rate of their internal AI programming platform grew too rapidly.
- Cisco warned that AI Agents consume a vastly larger amount of infrastructure compared to chatbots. According to Cisco, each employee in the future could be supported by anywhere from 10 to 1,000 AI Agents.
- Goldman Sachs forecasts that the volume of AI tokens used could increase 24-fold by 2030. The bank also warned that the surge in computing demand could exacerbate chip shortages over the next 12 to 18 months.
- Cost pressures could affect the growth prospects of AI companies like Anthropic and OpenAI ahead of major IPO plans.
- Data from OpenRouter shows that since the beginning of 2026, Chinese AI models have surpassed US models in terms of token usage volume.
- Chinese AI companies benefit from lower energy costs and more efficient models.
- Software company Workato stated that its AI costs surged 7-fold on just the first day after Anthropic switched to a token-based pricing model. Workato’s CIO described the situation saying: “We created a monster.” Instead of restricting usage, Workato is encouraging employees to switch to older and cheaper AI models.
- Amazon previously had to warn employees not to use AI just to climb internal leaderboards.
- Meta has also adjusted its AI usage incentive programs since April.
- Microsoft, Amazon, and Google are currently developing tools to automatically route requests to the most appropriate model to reduce costs.
- Some businesses are even switching to open-source models running on their own infrastructure to cut down the bills paid to AI vendors.
- Nevertheless, many companies continue to invest in AI because they have committed to investors regarding productivity gains and operational efficiency.
- 📌 Conclusion: The AI debate is shifting from “whether to use AI or not” to “whether AI is worth the cost.” Following a phase of aggressive usage encouragement, many large corporations like Uber, Walmart, Amazon, and Cisco are beginning to impose limits or alter policies as AI bills rapidly climb. The shift to token-based pricing models has made many companies realize that the actual cost of AI Agents can be much higher than expected. In the context of Goldman Sachs forecasting a 24-fold increase in global token usage by 2030, managing AI costs is becoming one of the greatest corporate challenges in the phase of large-scale AI deployment.
Businesses panicked as AI bills skyrocket 7-fold as the era of “free” AI draws to a close
Related Posts
Contact
Email: info@vietmetric.vn
Address: No. 34, Alley 91, Tran Duy Hung Street, Yen Hoa Ward, Hanoi City
© 2026 Vietmetric

