IT Cost Optimization: The AI Funding Engine

Every era of enterprise technology has had a defining investment priority. Something that bent budgets, architectures and roadmaps toward itself. In the late 1990s, it was the web. Businesses raced to put themselves online because the alternative was being left behind, then came cloud, which moved from being a technology initiative to an expectation for almost every CIO. Today, that priority is AI.

The scale of the shift is significant. The global AI market is expected to exceed US$1.42 trillion by 2032, growing at nearly 15% annually. AI is increasingly taking up a larger share of technology budgets, but the more important question is not simply how much organizations are spending on AI. It is where that money is coming from and whether the value being created is keeping pace with the investment.

The value gap is creating a funding challenge.

The infrastructure required to support AI is substantial. McKinsey estimates that around US$6.7 trillion in data-center investment could be required globally by 2030 to keep pace with demand for compute, with AI-capable data centers accounting for a significant share of that investment.

Yet the returns from AI are still developing.

McKinsey’s 2025 State of AI survey, covering 1,993 respondents across 105 countries, found that only 39% reported any impact from AI on enterprise-level EBIT. Among those reporting an impact, most said AI organizations’ EBIT.

This creates a clear tension. AI adoption is accelerating, but measurable enterprise value is not yet keeping pace with the scale of investment. That does not mean organizations are pulling back. Instead, they are changing where technology money goes.

A reshape, not a freeze.

BCG’s 2025 IT Spending Pulse points to this shift: spending on AI, GenAI, cloud and security is increasing, while more established categories such as server infrastructure, end-user devices, systems management and IT operations are facing greater pressure.

It represents a deliberate reallocation of technology capital.

The question for IT leaders is therefore changing.

It is no longer simply:

How do we spend less?

It is:

How do we free up more capital for the technologies and capabilities that will create the greatest value?

This is where cost optimization becomes strategic.

Where optimization lives

Much of the opportunity lies in familiar IT optimization levers.

Organizations can rightsize compute capacity instead of paying for infrastructure that is rarely used. They can move infrequently accessed data to lower-cost storage tiers, rationalize software licenses and eliminate overlapping tools. They can also review application portfolios, infrastructure utilization and cloud consumption to identify spend that no longer delivers sufficient value.

In the past, optimization was often associated with reducing the IT budget. In the AI era, the objective is increasingly to release capital from lower-value or underutilized technology spend and redeploy it towards higher-priority investments.

That means optimization needs to be connected directly to the organization’s technology investment strategy.

The next frontier is AI itself.

As AI moves from experimentation into everyday production, organizations need to manage the cost of AI workloads with the same discipline they have historically applied to infrastructure and cloud. This is where AI FinOps is emerging.

AI introduces new cost drivers that traditional IT financial management was not designed to manage. GPU and accelerator utilization, model selection, training and inference workloads, token consumption and workload frequency can all materially affect the economics of an AI application.

An idle compute cluster can represent significant wasted expenditure. The cost of every inference call can accumulate rapidly as an AI application scales. A model that performs well technically may still be economically inefficient if it consumes significantly more compute than an alternative.

This creates a new set of questions for technology leaders:

How much is being spent on training versus inference?
Which models and workloads are consuming the most compute?
How efficiently are accelerators being utilized?
What is the cost of serving an AI application at scale?
When should an organization use a larger, more capable model — and when is a smaller model sufficient?
What is the cost of an AI workload relative to the business outcome it produces?

These questions move AI cost management beyond traditional infrastructure accounting.

The goal is no longer to understand the technology bill. It is to understand the unit economics of AI.

From cost per workload to cost per outcome

This is where AI FinOps can become more strategic.

Traditional technology optimization often focuses on metrics such as infrastructure utilization, license costs or cloud consumption. AI creates an opportunity to connect technology consumption much more directly to business outcomes.

For example, organizations can begin measuring the cost of AI-enabled customer interactions, automated transactions, generated output or business processes alongside the value that these activities create.

This shifts the conversation from:

“How much does this AI workload cost?”

to:

“What business outcome are we getting for every unit of AI spend?”

That distinction matters because AI usage can scale quickly. Without visibility into its underlying economics, a successful AI application can become expensive.

Building cost visibility into AI architecture and operating models, therefore, needs to happen before AI workloads scale significantly, not after costs become difficult to control.

Optimization becomes an AI funding strategy.

The implication for CIOs and CFOs is clear.

AI does not necessarily require an ever-expanding technology budget. It requires a different approach to allocating technology capital.

One, organizations need to view the technology estate as a portfolio of investments. Some areas will continue to require funding. Others may be candidates for consolidation, modernisation or reduction. The capital released through those decisions can then be directed towards AI initiatives with a clear business case.

Two, AI investments need their own cost discipline. Organizations must understand what they are spending, how efficiently workloads are running and whether the resulting business value justifies continued investment.

This creates a two-sided optimization agenda:

Optimize the existing IT estate to fund AI.

Optimize AI itself to ensure that the investment delivers value.

Both are necessary.

Fund the future intelligently.

AI will continue to absorb a growing share of technology investment as it becomes embedded across business functions. The answer is not to constrain that investment simply because returns are still emerging.

It is to become more deliberate about where technology capital is allocated.

The organizations that gain the most from AI will not necessarily be those that spend the most. They will be those who can continually redirect capital towards the highest-value opportunities while maintaining visibility into the economics of what they are already funding.

That makes IT cost optimization more than a cost-management exercise.

It becomes the funding mechanism that allows organizations to invest in AI today, while building the financial discipline needed to make that investment pay off tomorrow.

References