The rapid adoption of artificial intelligence has created a new, uncomfortable problem for enterprises: nobody really knows how to budget for it. According to a recent report from No Jitter, the financial planning around AI initiatives is in a state of flux, leaving CFOs and IT leaders scrambling to allocate resources without a clear playbook.

This uncertainty comes as no surprise to industry watchers. The pace of AI innovation has outstripped traditional budgeting cycles, and the cost structures for AI projects—from compute resources to model training to ongoing maintenance—are wildly unpredictable. As a result, companies are finding that their carefully crafted budgets are often obsolete within months, if not weeks.

The Root of the Problem: Unpredictable Costs

One of the biggest challenges is that AI costs are not linear. The expenses associated with training a large language model, for instance, can scale exponentially with data size and model complexity. Cloud computing bills for AI workloads can balloon overnight as models are deployed to production and user demand spikes. No two AI projects have the same cost trajectory, and this variability makes it nearly impossible to forecast with confidence.

Furthermore, the human element cannot be ignored. Hiring skilled AI engineers and data scientists commands a premium, and the talent pool is shallow. Budgeting for salaries, training, and retention is as much an art as a science. Add in the cost of data acquisition, cleaning, and labeling, and the financial picture becomes even murkier.

"Nobody knows how to budget for AI," the report states, capturing the sentiment of many IT leaders who are navigating this uncharted territory.

Impact on IT and Procurement

The consequences of this budgeting blind spot are already being felt across organizations. IT departments are being asked to support AI initiatives without a clear understanding of the total cost of ownership. Procurement teams are struggling to negotiate contracts with cloud providers and AI vendors when they can't predict usage patterns. This leads to either over-provisioning (and wasted spend) or under-provisioning (and performance bottlenecks).

Moreover, the lack of clarity is stalling decision-making. Executives are hesitant to greenlight AI projects when they can't get a reliable ROI estimate. This is particularly true for generative AI tools, which are still relatively new and unproven in many business contexts. The result is a paradox: AI is both a top priority and a financial black box.

What Companies Are Doing Wrong

  • Using traditional IT budgeting models that don't account for AI's unique cost drivers.
  • Ignoring the cost of experimentation—many AI projects fail, and that failure has a price.
  • Failing to build in contingency for unexpected spikes in compute or data costs.

Toward a More Agile Approach

Experts suggest that enterprises need to adopt a more agile, iterative approach to AI budgeting. Instead of setting a fixed annual budget, organizations should consider rolling forecasts or quarterly reviews that can adapt to changing circumstances. This allows for greater flexibility in shifting funds to projects that show promise and cutting those that don't.

Another key recommendation is to invest in cost monitoring and optimization tools that provide real-time visibility into AI spending. These tools can help identify inefficiencies, such as underutilized GPUs or redundant data storage, and enable teams to adjust on the fly. By treating AI budgeting as a dynamic process rather than a static exercise, companies can better manage risk and maximize value.

Collaboration between finance and technology teams is also critical. CFOs and CIOs need to speak the same language, with finance understanding the technical constraints and IT appreciating the fiscal realities. This cross-functional dialogue can lead to more realistic budgets and better outcomes.

Key Takeaways

The challenge of budgeting for AI is not going away anytime soon. As the technology evolves, so too will the financial models that support it. Enterprises that embrace flexibility, invest in visibility, and foster collaboration will be better positioned to navigate this uncertainty.

  • AI budgeting is inherently unpredictable due to variable costs and rapid innovation.
  • Traditional budgeting methods are insufficient; agile, rolling forecasts are recommended.
  • Real-time cost monitoring and cross-team collaboration are essential for managing AI spend.

For now, the industry is still learning how to budget for AI—and that's okay. The key is to acknowledge the uncertainty and build systems that can adapt.