.NEXT 2026 - AI cost management and FinOps top of mind for digital leaders

The business reality of paying for enterprise AI usage was behind much of the hype at Nutanix .NEXT 2026 conference in Chicago, USA. Digital leaders in financial and professional services as well as hospitality, openly discussed the importance of using costing models and analytics to ensure token generation and usage doesn’t break the bank. The Cloud FinOps model was discussed as the most relevant available to digital leaders at present.

As AI and, in particular, agentic AI begin to change the operating model of organizations, digital leaders are once again having to address a new costing model for the computing power consumed. This must reflect not only the cost of operating AI tools and agents, but also be measurable against business metrics. Costs can increase if they deliver an improvement in service and therefore customers are retained, for example. But these are early days for enterprise AI, and many digital leaders are still grappling with how to align those costs with the bottom line of the firm.

A digital leader in our network recently revealed the reality:

For the token costs, the market is changing so rapidly. It is not just in-puts and out-outs, it is also about the caching. Engineers need to see what they are consuming in terms of tokens, otherwise you may get some nasty surprises.

Other digital leaders have shared that a developer used 55,000 tokens and created a bill of $100,000. That CIO says they had a rough meeting with the CFO.

Cloud FinOps (financial operations) is a model that is far from new and hasn’t had the take-up or championing it deserves. Speaking to Nutanix executives and attending CIOs, it was clear that this methodology is probably the best available to digital leaders at present, and they expect it to rise in importance.  

FinOps was created to enable business lines to share their cloud computing budget responsibility, and in an age of generative AI, inflation, and a climate emergency, that responsibility has never been more important. As we have discussed before, the utility nature of the cloud is both its strength and its weakness. Powerful compute is available with the same ease as turning a tap spigot or the flick of a light switch. Just as easy access to energy and water has led to ignorance about the impact these have on the natural environment and the subsequent damage to the natural ecosystem, so too has the utility of the cloud led to a disconnection between applications or storage used and the business implications.

AI has the same utility approach and set of risks.

Clarity

Digital leaders at .NEXT 2026 agreed that they need to provide a clear view of the costs of AI adoption now so that businesses are aware how and when it impacts the bottom line. Greg Lowe, CIO for hospitality provider Atlantis Bahamas and former CIO of Boyd Gaming says:

You have to show what is being done. Cloud is an easy conversation with people when you talk about keeping the business always on, speed to market, and all the innovative things its abilities enable. But you have to show the costs. That has to be available, and people want to see it.

Debojyoti Dutta, Nutanix Chief AI Officer, adds:

I feel there will be a bigger need for FinOps. You can get business gains from AI, but you have to account for how much you are spending on a daily and weekly basis. There are already firms that give their employees daily token credits. They are rationing it.

As he said in our previous article:

With technologies like Open Claw, you can generate six to seven million tokens as an engineer, writing an agent. Then multiply that by six to seven thousand employees, and the demand is exponential.

Dutta says CIOs will need data and platforms to access and interrogate the usage of AI with a clear eye on the bottom line. He adds:

When you do continuous operations, you need data to operate, so observability will become really important.

Tim Conners, CTO with global law firm Simpson, Thatcher & Bartlett LLP, added:

I wonder who's gonna be the first one to come out with a model to be able to measure the token consumption, because otherwise, it's gonna be the Wild West again.

Cloudy experience

Although Cloud FinOps hasn’t had the adoption and attention it probably deserves, digital leaders have learned from their cloud experiences.

Conners of Simpson, Thatcher & Bartlett LLP says peers do not want to be burned by nasty surprises. He adds that using FinOps has enabled him in this and previous roles to make sure there are not, in his words, too many zeros on the end of the bill. But more importantly, he says it is the ability to know and record your consumption and then have the tools to manage services up and down.

Cloud FinOps provides you with a decision tree component so you can see if a service belongs in the cloud (or uses AI). If it belongs there, is it a microservice? Is it something we can control, where it is not a 24/7, round-the-clock service? We have adopted an eyes-wide-open understanding of what the cost is. There are no surprises.

In financial services, Brandon Shaw, VP and Head of Technology Services for Western Union, said in a Nutanix customer panel discussion:

The cloud bug meant there was a push to put things there. Now there is a need to push the right things over to the cloud.

You can see in some quarters a similar rush to move workloads to agentic AI that may not deliver a real cost saving. Analyzing the workload and how that workload marries to the bottom line is an essential skill of CIOs and will be tested by the AI frenzy. CIOs must embrace AI, as it is a significant societal and business system change, but not at the cost of the bottom line. Conners says this plays to the other key role of digital leaders, making the business more effective:

Wherever you have waste, you can consolidate and bring down your overhead.

He says optimizing the estate will be critical, especially in the face of shortages:

If you don't have the capacity, and you're putting in an order now, you're not seeing it for three to four months.