AI cost surge prompts CFOs to tighten budgets.
AI budgets overshoot plans.
Multiple surveys of finance leaders in recent weeks reveal that AI projects are running far over budget, leading CFOs to crack down on spending. In a new Q3 survey of 2,000 CFOs and finance executives across the US and Europe, 83% said that usage-based AI services had pushed their costs beyond expectations ([1]). Unpredictable pricing models, where cloud AI vendors charge by tokens or usage, have made it difficult for businesses to forecast expenses ([2]). In many firms, AI budgets that were set early in the year have been exhausted long before year-end.
Finance chiefs are responding by tightening control of AI investments. Many companies are centralising AI spend under the CFO’s oversight and introducing “stage-gate” funding that releases money only when projects hit agreed performance milestones ([3]). According to Deloitte, 25% of large companies now require a formal business case with a quantified return on investment for new technology projects ([4]), though roughly another quarter admit they still lack any formal process to gauge the value of AI initiatives ([5]).
The days of blank-cheque AI spending appear to be ending. Over the summer, Uber discovered it had burned through its entire 2026 AI budget by April, just a third of the way into the year ([6]). Such experiences have become cautionary tales for peers. Finance leaders have begun reining in unbridled experimentation and scrutinising AI spending as operating costs mount. Even so, surveys show CFOs remain convinced of AI’s potential: despite cost overruns, finance executives on average rate their confidence in AI’s business value at 8.2 out of 10 ([7]), suggesting they are containing costs, not abandoning the technology.
Elusive returns on investment.
While AI spending has surged, tangible returns remain scarce for many organisations. A global survey of 4,454 chief executives by PwC found that 56% have yet to see any measurable financial benefits from their AI initiatives ([1]). Similarly, Deloitte reported that proving ROI is now the number one barrier to expanding enterprise AI programmes, cited by 54% of finance leaders surveyed ([2]). These findings underscore a hard truth: despite widespread AI adoption, more than half of companies are not experiencing clear bottom-line gains from their investments.
A key part of the problem is that many companies have struggled to even measure AI’s impact. Gartner notes that 84% of CFOs do not have a reliable way to quantify AI’s return on investment ([3]). In a recent U.S. Bank study of 1,000 senior finance leaders, respondents said they track ROI on just 41% of their AI projects, and only around 47% of those measured initiatives are delivering positive returns ([4]). In other words, barely one in five AI projects has proven to create real value so far. As one analyst observed, "AI value erosion is often not due to technology failure, but rather cost creep due to human factors" ([5]), such as implementing AI without clear metrics or cost controls.
Without defined baselines and accountability, the benefits of AI can remain invisible. Many firms rushed to invest in AI solutions over the past two years without first establishing how success would be measured. This lack of ROI planning means efficiency gains often go unconverted into profit ([6]) ([7]). Now, faced with greater board scrutiny, CFOs are urgently trying to retrofit measurement frameworks and financial discipline onto their AI projects. The goal is to distinguish initiatives that create real business value from those that merely generate cost.
Finance discipline uncovers real gains.
There are signs that when AI is tied to specific business metrics and overseen by finance, it can deliver concrete returns. Deloitte’s global Finance Trends survey finds 54% of finance leaders are now in charge of AI and technology investment decisions ([1]), integrating ROI accountability into tech deployment. IBM’s new "AI-first" CFO research goes further, reporting that companies with finance chiefs who treat AI as a strategic priority achieved 23% higher revenue growth than their peers ([2]).
Direct success stories are emerging where AI initiatives have a clear owner and measured outcomes. For example, a major insurance firm’s CFO reported using AI-driven virtual assistants in its call centres to handle routine customer queries, reducing call volumes by 20% and saving over $80 million in annual operating costs ([3]). Similarly, SpaceX, which built a large in-house AI supercomputing centre, has quickly turned that investment into a revenue stream. CFO Bret Johnsen noted in August that selling AI computing services to other companies has resulted in a payback period of under one year for SpaceX’s AI infrastructure spend ([4]). Just before its summer IPO, SpaceX signed multi-billion-dollar cloud deals with Google and Anthropic, giving the company a rapid return on its AI capital outlay ([5]).
These examples illustrate how value from AI can be realised when projects are approached with a disciplined business-case mindset. In each case, an accountable leader focused on a specific metric, whether customer service efficiency or data centre monetisation, and tracked progress against a pre-set baseline. For most organisations, such clarity is still the exception rather than the rule. But as CFOs demand evidence of outcomes before writing the next cheque, the hope is that AI investments will shift from open-ended experimentation to targeted initiatives that contribute tangible results.