The Costliest Employee in Tech May No Longer Be Human

For decades, technology companies budgeted around talent. Salaries dominated engineering costs, while software tools remained relatively cheap overhead. In 2026, that assumption is beginning to break.

A fully equipped AI coding stack now costs between ₹40,000 and ₹50,000 per month for a single engineer. Cursor, Claude Code, premium model access, and API usage can collectively cost $500- $600 (~₹42,000- ₹51,000) per month before companies pay the person using them.

A fresher at major IT services firms such as TCS, Infosys, and Wipro typically earns between ₹3.5 lakh and ₹8 lakh annually. For companies aggressively deploying AI coding tools, the software seat is rapidly becoming as expensive as the employee occupying it.

"The subscription cost is so high that it sometimes exceeds the cost of engineers themselves," Soumitra Mishra, Head of Academics at upskilling startup Newton School of Technology, tells AIM. He stated that the engineer is actually an employee of these AI tools, which he considers to be a very unusual situation.

The Bill That Broke the Budget

Last month, Uber's Chief Technology Officer Praveen Neppalli Naga confirmed to The Information that the company had burned through its entire 2026 AI budget before the second quarter began. The cause was not a failed experiment. It was a successful one.

Uber had rolled out Anthropic's Claude Code to its engineering team, roughly 5,000 engineers, in December 2025. By February, usage had more than doubled.

By April, 95% of engineers were using AI tools monthly. Approximately 70% of committed code was AI-generated. Around 11% of live backend code updates—covering everything from ride-matching to dynamic pricing—were written entirely by AI agents. This was up from under 1% just three months earlier. Monthly spend per engineer ranged from $150 to $250 on average. Heavy users spent up to $2,000 a month.

Similarly, at Nvidia, Bryan Catanzaro, vice president of applied deep learning, told Axios that for his team, the cost of compute far exceeds the cost of employees.

“Salaries in India are still relatively lower, that’s why people are feeling the heat where token costs per top engineer are going as high as ₹50,000- ₹1,00,000 a month,” says Raghav Gupta, CEO and Founder of Futurense.  

He adds that with AI majors OpenAI and Anthropic, expected to go public this year, the token costs may double or even triple, which will intensify the issue.

According to Gartner, global IT spending will hit $6.31 trillion in 2026, up 13.5% from 2025, with AI infrastructure and software subscriptions as the primary drivers.

A Structural Flip

For most of the history of knowledge work, the human was the expensive, scarce input. The tools were fixed assets: a laptop, a development environment, and a code library subscription. 

That equation has been inverted. Cursor's six plans now range from free to $200 per month for individual users; the business tier is $40 per seat per month, and enterprise contracts are custom-priced from there.

GitHub Copilot, its main competitor, charges individuals between $10 and $39 per month, depending on the plan, and is structured around usage-based limits for premium requests. 

Starting June 1, GitHub moved Copilot to a usage-based billing model, replacing its traditional system with a new token-consumption model.

Now add Claude Code, whose token-based pricing scales with usage and can reach thousands of dollars monthly for heavy users, and the per-engineer tool bill at an AI-forward company can exceed ₹1 lakh a month, comfortably above a junior Indian engineer's take-home pay.

Companies are at different levels of maturity in adopting GenAI and Agentic AI solutions at scale. Vijay Gopalakrishnan, Partner at Deloitte India, says, "Whoever embarks on the GenAI and Agentic AI journey involving tokens should be aware of the possibility of token costs spiralling out of control.” He notes that they must work with their teams and vendors to establish guardrails around the right processes, user access norms, design, governance, and LLM/Agent Ops, including FinOps, to monitor and manage costs, including token costs."

He adds that there are few instances where such guardrails were not implemented, leading to token costs becoming an issue at least temporarily. Companies are aware of this and are acting on it.

The Job That Remains

At Uber, the CTO's stated long-term vision is agentic software engineering, in which AI systems independently handle the full lifecycle of coding, testing, and deployment, with human supervisory agents auditing the output.

Hence, the premium in 2026 is not on writing code. It is about knowing when AI-generated code is wrong. That requires a different kind of knowledge: understanding enough about how large language models fail to catch their characteristic errors, confident-sounding hallucinations, subtle mismatches between what was asked and what was built, and security vulnerabilities introduced by autocomplete.

Industry data confirms the pattern. Workers who demonstrably apply AI engineering skills command a 56% wage premium over peers in equivalent non-AI roles, according to PwC’s 2025 Global AI Jobs Barometer.

As reported by AIM on May 20, the demand for Forward Deployed Engineers, operating on the front lines, embedded directly within a client’s organisation, has risen nearly 800% over the past year alone, with companies racing to find such talent.

These are not simply gradual changes in the role of a software engineer. Instead, they are distinct jobs with different requirements that just happen to carry the same title as previous roles.