Owen Lovelock

Business Development Director – Integres Software Solutions

How a new generation of platforms is making AI cost transparency achievable 

Over the past two articles, I’ve covered the scale of the AI investment problem. Firstly, addressing some of the reasons why 80% of projects are failing to deliver intended value and then in Part 2, looking at why AI costs are specifically hard to see: the infrastructure complexity, the GPU economics, the distributed workloads that don’t map neatly onto traditional cost management tools.

In this final part, I want to get practical. Because the good news is that the tooling to address this has matured considerably, and a small number of platforms are doing genuinely impressive work in this space.

Why this matters now 

Enterprise AI spending reached $307 billion in 2025 and is projected to reach $632 billion by 2028. At that scale, the financial discipline around AI investment isn’t optional; it’s an opportunity to step ahead of the competition.

The returns from getting it right are real. Organisations that implement intelligent cost optimisation across their AI infrastructure are achieving cost reductions in the range of 30–50%. Apptio’s own benchmarking suggests that organisations using their platform can reduce cloud unit costs by 30% or more and increase commitment coverage to over 90%.

These aren’t marginal improvements; they’re material changes to the economics of AI delivery.

Three layers of visibility and why you need all of them 

Meaningful AI cost transparency requires visibility at three different levels. The tools that are emerging as category leaders each address a distinct layer, and together they form a coherent picture of what your AI programme is actually costing. 

Layer 1: Strategic business-level TCO – Apptio TCO of AI 

At the highest level, the question isn’t just “what did we spend?” but “what did we spend, on what, and is it generating value relative to the business outcomes we care about?”

IBM Apptio’s TCO of AI capability is built for this conversation. It provides a structured framework for aggregating the full cost of AI initiatives across infrastructure, people and platform spend, and mapping that back to business value. For CIOs, finance leaders, and programme boards, it creates the single view that allows AI investments to be evaluated on a like-for-like basis and prioritised accordingly.

This is the layer that enables the conversation most organisations are missing: not just a cloud bill, but a genuine total cost of ownership picture that’s legible to the people making investment decisions. 

Layer 2: Kubernetes-level workload attribution – Kubecost 

The majority of AI workloads now run on Kubernetes, and Kubernetes creates a specific cost visibility challenge. Multiple workloads share the same infrastructure; compute, memory and GPU resources are dynamically allocated; and standard cloud billing gives you no way to see what individual pipelines, models or teams are actually consuming.

Kubecost solves this at the workload level. It provides real-time cost visibility down to the namespace, deployment and even the individual pod, so you can see precisely what a model training job is costing as it runs, which team’s workloads are responsible for which spend, and where GPU resources are being wasted on idle or oversized containers.

For engineering and platform teams, this is the granularity that drives accountability. When a team can see the financial impact of their infrastructure choices in real time, the conversations about right-sizing and efficiency get a lot more concrete. Kubecost 3.0 now includes advanced GPU monitoring via NVIDIA DCGM, making it one of the first platforms with genuine visibility into the GPU utilisation patterns that drive so much AI cost variability.

Layer 3: Continuous resource optimisation — IBM Turbonomic 

Visibility tells you what’s happening. Turbonomic goes a step further: it uses AI-driven analysis to continuously optimise how workloads consume resources, in real time, across hybrid and multi-cloud environments.

For AI deployments, this means intelligent right-sizing of the infrastructure supporting models – automatically matching resource allocation to actual demand, rather than leaving overprovisioned capacity running because nobody got around to scaling it down. In environments where GPU compute runs to thousands of pounds per day, automated continuous optimisation has material financial, time and resource impact. 

The recent integration between IBM Turbonomic and IBM Kubecost is particularly significant: it brings together real-time cost visibility (Kubecost) with intelligent resource action (Turbonomic), creating a feedback loop where cost insights directly drive optimisation decisions at the workload level.

The picture they create together 

Used in combination, these three platforms address the full cost visibility gap: 

• Turbonomic continuously optimises resource consumption at the infrastructure layer.
• Kubecost attributes that consumption in real time to specific workloads, teams and projects.
• Apptio rolls everything up into a business-level TCO view that connects AI spending to business outcomes.

The result is the end-to-end financial transparency that makes fail-fast genuinely achievable as an operational capability. When you can see what something costs, attribute it accurately, and act on that information in near real time, the decisions about where to invest and what to stop become straightforward.

Where to start 

If you’re exploring how to bring greater financial clarity to your AI programme, the honest starting point is the visibility layer. Understanding what you’re currently able to see, where the gaps are, and what decisions you’re making (or not making) as a result is crucial.

The platforms are there. The frameworks exist. The question is whether your organisation is ready to treat AI cost transparency as a strategic priority – not a finance team problem.

I’d love to continue this conversation. We’re bringing together a group of technology and finance leaders in Manchester this June to dig into exactly these challenges in a practical, peer-to-peer setting. If you’d like to be part of that, register your interest here.

In the meantime, if you would like to request more information about how Integres Software Solutions can help your organisation achieve greater value from your AI and Cloud investments through transparency, optimisation and automation, please submit the form below or email contact@techstories.ai.