The IT services landscape is shifting faster than most analysts predicted. Just a year ago, the consensus was that generative AI would slowly augment traditional software development over a decade. Now, that timeline has aggressively accelerated.
When Infosys CEO Salil Parekh recently took the stage at the Bloomberg New Economy Forum 2026, he dropped a metric that should make everyone in the enterprise tech space sit up: AI services are on track to drive nearly half of the company’s total work within two to three years.
Right now, this segment accounts for just over eight percent of their revenue. But the sequential double-digit growth tells a much larger story about how global enterprises are fundamentally rewiring their operations from the ground up.
The Economics of Generative IT
This isn’t just about selling a new software package or bolting a chatbot onto an old website. What we are witnessing is the industrialization of AI agents and automated coding workflows, proving once again that AI won’t replace your team — but it will replace your workflow.
Traditional IT modernization used to mean migrating old databases to the cloud over a painstaking five-year period. Today, it means deploying intelligent systems that can actively read, rewrite, and optimize legacy codebases with minimal human intervention.
Infosys already has roughly 80,000 of its three lakh employees working directly on AI-driven coding tools and internal deployments. That massive scale requires serious talent, explaining their push to onboard 20,000 fresh graduates this financial year specifically trained to manage this technological transition for clients.
According to a recent industry report covering these strategic moves at People Matters, the sheer pace of this adoption could see AI services command 20 to 30 percent of the firm’s workload in just a few short quarters.
The mechanics of this shift are fascinating. Rather than billing solely for human hours spent writing raw code, the service model is evolving into AI orchestration. Teams build custom guardrails around models, feed them client-specific data, and deploy agents that handle the heavy lifting.
The human engineers act as reviewers and strategic directors rather than typists. This explains why Parekh remains so optimistic about the broader Indian IT sector. The lingering fear was that AI would cannibalize coding jobs, but the reality taking shape is that it creates a massive new market for tech services capable of managing these complex systems safely at an enterprise scale.
Navigating Infrastructure and Token Economics
Aggressive scaling brings an entirely new set of headaches, mostly centered around computing costs and hardware longevity. As the initial hype settles, enterprises are getting remarkably practical about how they spend their IT budgets. Instead of defaulting to the most expensive frontier models for every single task, clients are exploring how to reduce AI API costs.
They are actively hunting for lower-cost tokens and leaning into cheaper, open-weight models that can handle routine tasks just as effectively as the massive closed systems.
This shift fundamentally alters the hardware equation. Parekh correctly identified that the physical infrastructure specifically the massively expensive chips powering these data centers faces a precarious future. The software is advancing faster than the concrete can dry on the server farms.
If new foundation architectures suddenly make inference drastically cheaper, the billions currently being poured into today’s specific chip clusters might not yield the long-term returns investors expect.
Throw in the complexities of strict sovereign data laws and the growing social friction around aggressive AI expansion in Western markets, and the roadmap gets incredibly complicated. The demand is undeniably fantastic today. Yet, balancing that surging enterprise appetite against tomorrow’s inevitable hardware depreciation makes this current growth phase as volatile as it is lucrative.
