AI Spend Is Now the Biggest Cloud Cost Blind Spot

Week of 26 October 2026 | Compiled by Team Paringa

AI has become the fastest-growing and least governed line in the cloud bill. FinOps teams are tracking it, but ownership, visibility and enforcement are lagging.

What Happened This Week

🔴 Harness: AI spend is up, owners are missing, surprises are routine

Harness's 2026 State of AI in FinOps study, based on a May to June 2026 survey of 700 engineering leaders and practitioners at organisations with 1,000 or more employees in the US, UK, France, Germany and India, paints a picture of spend outrunning control. 67% spend more than $250K a month on AI, and 20% have crossed $1M a month. 80% say AI spend rose over the past six months and 78% expect it to keep climbing.

The governance gap is stark. 52% say there is no clear, dedicated AI cost owner, with accountability split across Platform and DevOps (30%), dedicated FinOps (27%), Finance (23%) and Engineering (19%). 72% had an unexpected AI cost spike or surprise bill in the past 12 months, and one in three were hit more than once. 79% need a day or more to trace a spike, 32% need a week, and 8% never reliably find the cause. Respondents estimate 26% of AI spend is wasted.

Policy exists but is not enforced: 73% have AI cost policies, yet only 47% fully enforce them. 56% describe AI budgeting as guesswork, only 13% report basic cost visibility, and 57% say their organisation encourages what the report calls tokenmaxxing, maximising AI use regardless of value. This sits on top of Flexera's State of the Cloud 2026, which put wasted IaaS and PaaS spend at 29%, up from 27% and reportedly the first increase in five years, with AI workloads and complexity blamed. Note the Harness findings are a vendor survey, so treat the figures as indicative.

Sources: Harness | Shattered, summarising Flexera

The headline is not that AI is expensive, it is that nobody owns the bill. A 52% no-owner rate means most organisations cannot answer a simple question: who gets the call when spend spikes? Start with ownership and tagging at the workload and model level, then add anomaly alerts that fire in hours, not weeks. Tools like Atmoz exist to give that visibility, but the first step is a named owner and a monthly review cadence that is not quarterly.

🟡 Shadow AI subscriptions are quietly inflating software spend

Research from Marketing Signals, reported by Compare the Cloud on 31 August 2026, found spend on individually expensed AI subscriptions rose 108% year on year. These tools, typically $20 to $40 a month each, are bought on personal cards and claimed back, so they rarely appear in procurement or software budgets. The agency estimates unauthorised and forgotten AI subscriptions cost UK businesses about £63.7 million a year, within a range of £34.0 million to £112.1 million. The article frames this under marketing departments.

Between 50% and 71% of employees report using AI tools their employer has not approved, and Vertice data cited in the piece shows SaaS licence waste rising from 62% to 65% over the past year. In the agency's own audit, removing or replacing five tools cut monthly subscription costs from £1,875 to £713. This item is older than our usual window, but it is the freshest quantified view of shadow AI spend we found and it reinforces the Harness findings.

Sources: Compare the Cloud

Shadow AI is a cost problem before it is a security problem, and it hides in expense claims, not cloud invoices. Pull expense data and card feeds into the same view as cloud spend and you will usually find quick wins within a month. Pair it with an approved-tools list so people have a sanctioned route, otherwise the spend simply reappears elsewhere.

What Finance and Engineering Leaders Are Dealing With

FinOps has matured around compute and storage, but AI spend behaves differently: it is usage and token driven, it spikes without warning, and it is often bought outside central teams. The Harness data shows 98% of FinOps teams now track AI in some form according to G2 figures cited alongside it, yet few can attribute or enforce.

The next phase is less about dashboards and more about accountability: named owners, unit economics per AI feature, and guardrails built into engineering workflows so cost is considered at design time.

Three Things Worth Watching Next Week

1. Q4 budget resets — Expect finance teams to challenge AI line items as 2027 budgets are drafted, so anomaly and attribution data will matter.

2. Vendor responses to AI cost visibility — Cloud providers and FinOps platforms are racing to ship model-level cost tooling, so watch for announcements.

3. FinOps Foundation community output — Watch finops.org for new guidance on AI cost management and scope expansion beyond public cloud.

Questions? hello@paringa.ai

Paringa helps organisations bring cloud and AI spend under control without slowing teams down. We combine partner technology with hands-on FinOps practice so savings stick. | 26 October 2026 | © Paringa Ltd