Indian companies adopting artificial intelligence are facing rapidly rising cloud and infrastructure costs as enterprise spending shifts away from labour-heavy IT services towards GPUs, cloud subscriptions and AI platforms.
“The assumption was that companies could remove labour costs and replace them with cheaper AI infrastructure, but in many cases AI infrastructure spending is now growing faster than payroll savings,” said an analyst tracking enterprise AI spending trends who wished to be unnamed. “Firms are discovering that AI improves productivity, but not always enough to offset rising inference, cloud and subscription costs.”
Research firm Gartner said global AI spending is expected to rise 47% year-on-year to $2.59 trillion in 2026, with AI infrastructure accounting for more than 45% of total spending. At the same time, UnearthInsight said enterprise technology budgets are increasingly rotating “from people-hours to platform-subscriptions”, as companies spend more on cloud inference, AI agents and AI infrastructure.
The pressure is becoming more visible among Indian IT firms, banks, startups and large enterprises deploying AI tools across coding, customer support, workflow automation and internal operations.
According to UnearthInsight, cloud subscription growth in FY26 ranged between 19% and 27%, while IT services revenue growth slowed to between negative 2% and positive 4%.
“In many enterprises, labour savings from AI-led restructuring are being outweighed by rising spending on GPUs, cloud inference and enterprise AI subscriptions. Companies may save 5-10% on workforce costs in the near term, but AI infrastructure and model consumption costs are growing 20-30% annually, while revenue growth in core IT services businesses remains in low single digits. The result is that AI is improving productivity, but not yet delivering proportional profit expansion for most firms,” the AI analyst said.
An Avendus Capital report estimated rising AI adoption could drive deployment of 650,000 to 700,000 GPUs across Indian data centres over the next five years, creating a $23 billion infrastructure opportunity by 2030. The economics are becoming harder to manage because many Indian enterprise AI workloads still depend on US cloud providers and Nvidia-powered GPU systems priced in dollars.
Uber recently said it exhausted its annual budget for Claude Code, Anthropic’s AI coding assistant, by April as internal employee usage surged. Uber chief operating officer Andrew Macdonald said there was still no clear evidence that higher AI usage automatically translated into better products.
“That link is not there yet,” Macdonald said during a podcast discussion on enterprise AI spending. Further, the productivity challenge is becoming more visible across enterprises scaling AI aggressively.
A recent HCLTech report found that nearly 43% of major enterprise AI initiatives are expected to fail as enterprise leaders are expecting measurable returns from AI investments within 18 months.