Why Cheap Technology Can Become an Expensive Business Decision

A cheaper unit price does not always mean a cheaper bill. Gartner said on August 17, 2026 that AI inference costs per agentic workflow will rise more than fivefold through 2028, even as the price of each unit of AI processing, called a token, keeps falling.

The pattern is not limited to AI. Cloud computing and security tools show similar gaps between the price on the invoice and the cost of owning the technology.

The price you see versus the cost you carry

The purchase price is only one part of what a technology costs. Total cost of ownership also includes integration, growth in usage, the staff who run it, the controls that keep it safe, and the cost of switching away. A low entry price can hide all of these.

When cheaper units drive bigger bills

Gartner calls it the "inference paradox." It points to three trends. Foundation model economics are improving. That efficiency is being used to deploy more powerful and often more expensive models. And more sophisticated workflows use far more tokens than a simple chatbot.

Routing a task to an agentic reasoning model raises provider inference costs by at least five times compared with a basic chatbot exchange, Gartner said. Gartner also expects inference on a trillion-parameter language model to cost providers more than 90% less by 2030 than in 2025, according to reporting on its March forecast. The savings are real, but they do not automatically reach buyers' budgets.

"Product leaders cannot rely on more efficient token economics to rationalize AI costs," said Will Sommer, a senior director analyst at Gartner.

Gartner has also predicted that over 40% of agentic AI projects will be canceled by the end of 2027 because of escalating costs, unclear business value or inadequate risk controls. That is a forecast, based partly on a January 2025 poll of webinar attendees, not a measured outcome.

Cheap to start, costly to scale: the cloud

Cloud services make it easy to start small, which is part of their appeal. Scale is where the discipline is tested.

Flexera's 2026 State of the Cloud report, based on 753 respondents, estimated that 29% of spending on infrastructure and platform services is wasted. That was up from 27% a year earlier and the first rise in five years. Seventeen percent of respondents said they had exceeded their public cloud budgets over the past year.

The figures are self-reported estimates, not audits. Flexera linked the rise to growing cost complexity from AI and new cloud services.

Cheap tools outside oversight: security

Low-cost tools also carry risk when they spread without governance. IBM's 2026 Cost of a Data Breach report, published in July, put the global average cost of a breach at a record $4.99 million, up 12% from the year before. The US average was $11.5 million.

IBM found that incidents involving "shadow AI," meaning AI tools employees use without approval from IT or security teams, appeared in 43% of breached organizations, up from 20% a year earlier. Those breaches averaged $5.39 million. IBM also said AI-driven attacks rose 56% and added about $1 million to the average breach.

An AI tool that costs little or nothing to adopt can become expensive if it exposes data or draws a regulatory fine. About one in five shadow AI incidents involved a fine, according to summaries of the report.

Where the expensive decisions come from

These cases share three features, which is analysis rather than a finding from any single report. First, the cost scales with usage, not with the purchase price. Second, much of the cost sits in someone else's budget, such as security, finance or compliance. Third, it arrives late, after the decision has been made and the tool is embedded.

What buyers can ask before committing

• What happens to the bill at ten times the usage? Ask for a forecast, not just the entry price.
• Who owns monitoring? Assign someone to track spending and set alerts before launch.
• What does one successful outcome cost? Measure cost per outcome, not per unit.
• What are the exit costs? Check data export, contract terms and retraining effort.
• What data can this tool see? Set approval rules, since staff will find tools on their own.

What this does not mean

None of this argues against adopting new technology. Unit costs are falling, and many organizations get real value from AI and cloud services. The point is narrower: the number that matters is the total cost of the system, including how it behaves as use grows.

For buyers, a low price is a reason to look closely, not a reason to stop. The question is what the technology will cost once the whole business depends on it