Only 4% of Senior IT Budget Leaders Are Highly Satisfied With AI Cost and ROI Tracking

Nicus research found that only 4% of senior IT and technology budget leaders are highly satisfied with their ability to categorize and track AI usage, benefits, token-consumption costs, and returns on AI investment. The findings highlight a significant enterprise governance and financial-management gap as organizations increase spending on AI.

Analysis

This is weak standalone market-moving information, but it reinforces a developing enterprise-software bottleneck: AI budgets are moving from experimental spend to CFO-controlled unit economics. The vendors most exposed to this shift are not necessarily model providers; they are FinOps, cloud-cost management, IT service-management, and data-governance platforms that can attribute GPU, inference, storage, and labor costs to business units. Near term, the likely effect is slower conversion of loosely defined AI pilots into broad production deployments, particularly where ROI cannot be tied to revenue, labor savings, or customer retention.

For hyperscalers, the second-order risk is that poor workload attribution raises scrutiny of inference consumption and could lengthen enterprise commit cycles over the next 1-3 quarters. This is more relevant to Microsoft Azure and Google Cloud, where incremental AI usage is expected to support cloud reacceleration, than to Nvidia, whose demand is still primarily constrained by customer capex and supply; however, a sustained ROI gap becomes a 6-18 month risk to the durability of GPU-order growth after initial capacity build-outs. ServiceNow and IBM are potential beneficiaries if customers respond by formalizing AI governance and workflow measurement rather than cutting projects outright.

The consensus may be too focused on aggregate AI capex and insufficiently focused on the deployment layer. Enterprises can maintain infrastructure spending while reducing the number of user-facing applications, creating a divergence between compute demand and application-software monetization. The key falsifier is evidence in upcoming earnings calls that AI pilots are converting to paid production seats and measurable customer outcomes faster than budget governance is tightening.

Key Decisions for Investors

• No directional trade solely on this survey; treat it as a diligence flag ahead of MSFT, GOOGL, NOW, IBM, and SNOW earnings. Monitor disclosed AI revenue, cloud consumption growth, and commentary on pilot-to-production conversion over the next 1-3 months.
• Favor a 6-12 month relative-value expression: long NOW / short SNOW in equal dollar amounts, conditional on ServiceNow sustaining subscription-growth guidance while Snowflake does not show accelerating consumption from production AI workloads. The thesis is that governed workflow deployment monetizes earlier than open-ended data/AI experimentation; exit if SNOW product revenue reaccelerates by more than 300 bps for two consecutive quarters.
• For AI infrastructure exposure, retain NVDA but reduce incremental-beta additions if hyperscalers begin citing enterprise ROI scrutiny, utilization gaps, or delayed AI workload deployment. A practical risk trigger is a material cut to 2027 capex expectations from two or more major cloud buyers, which would challenge the post-buildout demand narrative.
• Watch private FinOps/ITFM vendors rather than forcing a public-equity proxy. An acquisition or strategic investment by ServiceNow, IBM, Oracle, or a hyperscaler would validate that AI cost attribution has become a required control plane and could create a tactical catalyst for the acquirer.