Unpacking the 18-Fold Increase
The headline figure from Gartner, discussed at its recent IT Symposium in Kochi, highlights a sharp turn in the economics of artificial intelligence. Two years ago, in late 2024, many companies were in the early stages of AI adoption, running smaller
pilot programs and experiments. Today, as we approach 2027, the ambition has shifted to enterprise-wide deployment of more complex systems, including generative and agentic AI. This leap from isolated experiments to deeply integrated business tools is the primary reason for the staggering cost increase. Gartner analyst Neha Kumar noted that while over half of organisations are seeing positive results from their AI initiatives, the cost to achieve those gains is forcing companies to be far more selective about their investments. It's no longer about whether AI can work, but whether it can work profitably.
The Soaring Price of Intelligence
Several key factors are driving this cost inflation. The first is the sheer price of computational power. Training and running large-scale AI models requires immense processing capabilities, primarily from specialised GPU chips. Building and maintaining the data centres that house this hardware is a multi-trillion-dollar global endeavour, with spending on AI infrastructure expected to hit nearly $1 trillion in 2026 alone. Another major driver is the move from simple conversational AI to more advanced 'agentic' systems. These are AIs designed to autonomously perform complex, multi-step tasks, which can consume 10 to 100 times more resources per session than a simple query. Finally, the cost of talent remains high, with skilled AI engineers and data scientists commanding premium salaries. Together, these elements—compute, data, talent, and energy—create a financial barrier that is far higher than it was just 24 months ago.
A Shift from Experimentation to Infrastructure
The 18-fold increase also reflects a fundamental change in how businesses view AI. In 2024, AI was often a software-level experiment. In 2026, it is an infrastructure-level commitment. Companies are no longer just buying AI-powered software; they are building foundational platforms. According to a Gartner CEO survey, 86% of chief executives see redesigning their operating models as crucial for generating value from AI. This means reworking entire workflows and business processes, which is a far more significant and costly undertaking than simply adopting a new tool. This shift transforms AI from a line item in a department's budget into a major capital expenditure, similar to building a new factory or data centre. The scale of this investment is massive, with some analysts warning that the revenue generated from AI may not be enough to justify the trillions being spent on infrastructure, creating risk for investors.
The Impact on Indian Businesses
For Indian companies, this trend presents both challenges and opportunities. The nation's massive IT services sector is at the forefront of helping global clients implement AI, driving demand for Indian tech talent. However, the rising costs mean that both service providers and their clients must be more strategic. Gartner's findings were discussed in Kochi, underscoring the relevance of this global trend to the Indian market. Indian enterprises looking to deploy AI must now grapple with the same 'build vs. buy' decisions as their global peers, but with a keen eye on the return on investment. The high cost of building proprietary models may lead many to rely on APIs from global tech giants, while others might focus on smaller, domain-specific models that are cheaper to train and run. Governance is another critical concern, with one survey finding only 14% of CIOs in the APAC region, including India, have adequate governance for agentic AI.
Navigating the New Cost Landscape
As costs spiral, a new focus on 'AI FinOps'—financial operations for AI—is emerging. The goal is to maximise value while controlling runaway spending. Businesses are being urged to implement better governance and cost-monitoring dashboards to track expenses in real time. Strategies include using smaller, more efficient models for simpler tasks—a practice known as model routing—instead of defaulting to the most powerful and expensive option. Other tactics involve optimising how data is handled to reduce processing fees, caching frequent queries to avoid redundant work, and setting hard spending limits for teams and projects. As one Gartner analyst put it, a cheaper AI model doesn't always lead to a cheaper business outcome. The key is to connect every AI investment to a clear business objective and measure its impact rigorously, ensuring that the productivity gains justify the now-stratospheric cost.
















