The cloud bill just became a board conversation

For years the cloud bill sat quietly in the middle of the IT budget, growing steadily but never dramatically. AI has ended that. Gartner forecasts worldwide AI spending will hit US$2.59 trillion in 2026, up 47 percent on last year, with AI infrastructure alone climbing from around US$976 billion to US$1.43 trillion. Most of that money is hyperscalers building capacity rather than enterprises buying directly. But capacity gets built to be sold, and the bill flows downhill.

It is already landing. A survey of 100 CFOs at midsize IT and SaaS firms, reported by CIO, found cloud is now the second-largest expense behind payroll, averaging 10 percent of revenue, with AI and machine learning workloads making up 22 percent of cloud costs. Those AI costs are also the least predictable part of the bill, with training spikes and usage-driven inference making forecasting genuinely hard.

From movement to expectation

FinOps, once a niche enthusiasm, now looks like table stakes. The FinOps Foundation’s State of FinOps 2026 report, drawing on 1,192 practitioners representing more than US$83 billion in annual cloud spend, found 98 percent of teams now manage AI costs, up from 63 percent a year earlier. The discipline’s remit has spread well beyond public cloud: 90 percent of teams now manage or plan to manage SaaS, 64 percent software licensing, 57 percent private cloud and 48 percent the data centre. And it has moved up the org chart. CIO Dive reports that 78 percent of FinOps teams now report to the CTO or CIO, up from 61 percent in 2023. Cost management has become an engineering and architecture capability, not a finance chore.

Here is the uncomfortable bit: all that maturity has not stopped waste rising. Flexera’s 2026 State of the Cloud report found estimated wasted cloud spend ticked up to 29 percent, the first increase in five years, and points squarely at surging AI workloads with unpredictable usage patterns as the cause.

Repatriation, minus the hype

The loudest storyline of the year is workloads coming home. A Broadcom survey of 1,800 senior IT leaders, reported by The Register, found 83 percent of enterprises considering repatriation, up from 69 percent in 2025, and the share running production AI inference primarily in public cloud falling from 56 to 41 percent. Notably, cost overtook security as the top public cloud concern for the first time. Broadcom owns VMware and sells private cloud, so season that to taste. Flexera’s broader data suggests rebalancing rather than exodus: hybrid estates grew to 73 percent of organisations and cloud spending keeps climbing. Nobody credible is calling time on public cloud. They are calling time on unexamined defaults.

For New Zealand leaders there is a currency multiplier on all of this. Hyperscaler bills are priced in US dollars, and the Kiwi has been trading around 58 US cents, a few percent weaker than a year ago. The same invoice costs more in New Zealand dollars before your usage grows at all. Meanwhile vendors are openly passing their AI infrastructure costs through: Forrester’s survey of 2,600 decision-makers found software budgets rising as vendors lift prices or add usage charges, with 80 percent expecting data and software spend to increase.

The thread running through every one of these reports is that cost discipline has become a leadership capability rather than a procurement task. Forrester’s chief research officer put it plainly: “The organizations that outperform in 2027 won’t be those that spend the most on AI.” For NZ tech leaders heading into budget season with a soft dollar and hungry workloads, that is oddly good news. The advantage is not going to the biggest wallet. It is going to the sharpest one.