Something has quietly inverted in the economics of the modern corporation. For three decades, the marginal cost of knowing something inside a large organisation fell steadily toward zero, pushed down by cheap storage, cheap bandwidth and dashboards that multiplied faster than anyone could govern them. That curve has broken. Every answer an enterprise generates now carries a metered price, and boards are being asked, for the first time since the arrival of the data warehouse, to treat knowledge itself as a cost line with a run rate.
Intelligence has become a utility bill. Watch what actually happens when a business user poses a question to an AI system. Compute cycles are consumed, tokens are burnt, and retrieval layers reach across systems that were never designed to speak to one another. Behind all of it, a governance obligation must be satisfied before anyone can act on the output. The insight lands in seconds. The invoice lands monthly, and it does not shrink.
That invoice is arriving on desks with no established framework for reading it. A survey by Gartner found that 55 per cent of chief supply chain officers cannot clearly state the return on their AI investments, even as 67 per cent of supply chain digital investment now flows into AI. The pairing deserves a second read. Two-thirds of the money goes to a category whose payback more than half the buyers cannot articulate.
Capital markets have noticed. Investor tolerance for the AI narrative has thinned noticeably over recent quarters, and the questions surfacing on earnings calls have moved from what a company is building to what a company is booking. Ambition no longer clears the bar. Realised throughput does. For Indian enterprises carrying global delivery mandates, that scrutiny arrives twice over, once from domestic shareholders and once from clients in London and New York running the same arithmetic.
Where enterprises are struggling on the demand side, the state has moved decisively on the supply. The Union Budget 2026-27 introduced a tax holiday running to 2047 for eligible foreign cloud providers operating through India-based data centre infrastructure, a measure the government set against a global picture in which data centres accounted for more than a fifth of worldwide greenfield project value in 2025, with announced investment above 270 billion dollars, on UNCTAD figures cited by the Press Information Bureau. Layer on the IndiaAI Mission outlay of ₹10,371.92 crore cleared by the Cabinet, and the direction of travel reads clearly enough.
None of that changes the fact that the cost curve is actually bending upward. Cheap compute produces answers faster. It does not make them defensible, and in a regulated boardroom, a rapid answer nobody can trace is worth less than no answer at all.
Ask where the budget genuinely disappears, and the answer sits well below the model layer. Gartner has reported that 63 per cent of organisations either lack or are unsure they have the data management practices AI requires and predicts that through 2026 organisations will abandon 60 per cent of AI projects unsupported by AI-ready data. Those initiatives rarely die on the algorithm. They die on lineage nobody mapped, metadata nobody made active, a customer definition that means one thing in finance and something else in operations, and access policies drafted for quarterly reporting while the model consumes data by the hour.
The truth is the real cost of knowing. Not the query, but the proof standing behind it.
Enterprises that have absorbed the lesson are behaving differently. Governance is being treated as a revenue enabler rather than a compliance tax, a single source of intelligence is being insisted upon before any copilot goes live, and adoption is being measured instead of demos being applauded. Time to value has replaced novelty as the metric that survives the budget review, and the work increasingly begins by connecting and curating what already exists rather than replatforming a working estate.
There is a human ledger here too, and it is the one most often ignored. Every executive who receives an answer still has to decide whether to act on it, and that decision carries a private cost measured in reputation. Confidence is expensive. Trust is what makes the investment case close, and it is manufactured upstream, in the unglamorous machinery of quality, lineage, and ownership.
The organisations that emerge from this cycle ahead will not be the ones that spend the most on intelligence. They will be the ones who built a foundation solid enough to turn data into decisions at a price the business can defend, quarter after quarter, without flinching.
