The use of AI is expanding, and AI companies are bringing new and improved models to capture a share of this growing market. Earlier, these companies were spending billions of dollars of investors’ money on training, and deploying complex AI models, but now looking at alternative sources of revenue as funding from investors and banks is drying up. As an alternative, they have slowly started recovering costs from their customers, and corporations are now feeling most of this heat, as they are the biggest users.
For some corporations, the cost of deploying AI across their workflows has risen beyond their employees’ salaries. Recently, Uber CTO Praveen Neppalli Naga said they had used their full-year AI budget just a few months into 2026 due to the surging use of AI coding tools, particularly Claude Code by Anthropic.
Similar sentiments were echoed by Amos Bar-Joseph, CEO of getswan.com. In a LinkedIn post, he shared that their AI bill had reached $1,13,000 (around Rs 1,06,49,000) in a single month for a four-person team.
Even companies that work with other AI firms as infrastructure providers feel the same pressure. “For my team, the cost of compute is far beyond the cost of employees,” Bryan Catanzaro, vice president of applied deep learning at Nvidia, told Axios.
The rising cost of AI is emerging as a challenge for corporations worldwide. IT spending is expected to reach $6.31 trillion in 2026, up 13.5% from 2025, according to Gartner, covering everything from AI buildouts to subscription costs. Corporates will need to prove returns on AI spending over time to their investors and shareholders and demonstrate how rising AI costs are impacting overall operations.
Users feel the pressure as well
The impact is not just limited to corporations. Previously, Anthropic users have also complained that they are unable to get the maximum use out of Claude, as many have reported faster usage of their token limits, which they get by subscribing to a paid plan.
AI companies have also started keeping their more advanced and feature-rich models behind paywalls, which has impacted access to these features for regular users. A recent survey by Epoch AI highlighted that the best and most expensive AI tools, such as Anthropic Claude, are increasingly finding favour with richer users, whereas people with lower incomes are trying to make do with tools like Meta AI that are currently significantly inferior.
What next?
The technology is still improving each day, and AI companies are trying to minimise costs for their customers. Google has recently rolled out new TPUs that consume less power for training and deploying AI.
Similarly, OpenAI has just released its GPT-5.5 model, which OpenAI claims uses significantly fewer tokens to complete the same Codex tasks, making it more efficient as well as more capable.
Moving forward, the cost of using AI could decrease as companies introduce more efficient systems for running it.