AI investments are soaring - but where’s the ROI?

Enterprise AI adoption has never been higher, yet **returns on investment (ROI) remain stubbornly elusive**. Consulting firm McKinsey’s latest global survey of 1,719 organizations found that while 88% of companies now use AI in at least one business function, only **37% can point to any positive impact on earnings from their AI initiatives** ([1]). Tellingly, that 37% figure is essentially flat compared to a year earlier, underscoring a persistent gap between adoption and actual business value ([2]).

Perhaps more striking, a mere **6% of companies qualify as "AI high performers"** that attribute at least 5% of their enterprise-wide earnings to AI – the same tiny minority as last year ([3]). In plain terms, fewer than one in ten firms are achieving significant profit from AI at scale, even as nearly nine in ten are experimenting with the technology.

Nor is this ROI shortfall limited to one survey. In another large study of almost 6,000 executives across major economies, more than **80% reported no discernible boost to overall productivity or employment from their AI deployments** ([4]) – despite 69% of those companies already using some form of AI ([5]). Simply put, for the vast majority of organizations, early AI projects haven’t yet moved the needle on key performance metrics, falling far short of the transformative promises that fueled the spending spree.

Interestingly, this mismatch between expectations and outcomes hasn’t dampened enthusiasm. As one report notes, companies’ **conviction in AI is growing even faster than the immediate financial returns** they’ve seen so far ([6]). More executives than ever believe AI will reshape their business in the next few years and plan to keep increasing investment ([7]) – even if clear ROI remains a work in progress.

The lack of tangible ROI is putting pressure on C-suites to justify every AI dollar. **Finance chiefs in particular are losing patience with AI initiatives that can’t demonstrate real value**. In a recent survey, fully **92% of CFOs said they feel personal pressure to show that their AI investments are delivering a decent return** ([1]). Yet only 7% of those finance leaders believed their organizations prioritize careful AI governance over rapid adoption, reflecting a widespread “act now, figure out oversight later” approach ([2]).

Facing intensified scrutiny from boards and shareholders, CFOs are increasingly instituting **“ROI or no-go” rules for AI projects**. The **2026 Deloitte CFO AI Survey** found that 68% of CFOs will not approve further funding for AI without proof of ROI, and **42% have already cut pilot programs that failed to produce measurable results** ([3]). As one industry observer put it, this is a turning point: the era of blank-check AI projects is over, and **AI budgets must earn their keep or face the axe**.

Investors, too, are watching closely. After years of cheering on big AI spending, market sentiment is shifting to **favor companies that can translate AI investments into profits**. When Google’s parent Alphabet announced a massive increase in AI capital expenditures this summer, its stock price tumbled 7% on the news – erasing $294 billion in value – as investors balked at surging costs without clear returns ([4]). The message from both the boardroom and Wall Street is unmistakable: **show us the real ROI, or stop pouring in more cash**.

While business outcomes lag, the **costs of AI initiatives are exploding**. Gartner’s latest analysis projects that **global AI spending will reach $2.7 trillion in 2026, up 49.5% from 2025** ([1]). The ongoing build-out of AI capabilities – from cloud data centers to specialized chips – has been described as **“the largest infrastructure project humanity has ever undertaken”** ([2]). In other words, companies are pouring unprecedented capital into AI, treating it as a must-win race despite uncertain returns.

The financial strain of this AI arms race is already evident. Deploying cutting-edge models requires enormous cloud computing power, and the bills can add up fast. In one dramatic example, an enterprise that gave employees unfettered access to AI tools **ended up with a $500 million cloud bill in a single month** – only discovering the error when the invoice arrived ([3]). Even without such extreme mishaps, the pursuit of advanced AI has driven up operating costs. Earlier this year, Alphabet projected over $190 billion in 2026 capital spending to expand its AI infrastructure, contributing to a **–$5.9 billion free cash flow in Q2** and spooking investors ([4]).

These economics have some leaders rethinking the **build-vs-buy** equation for AI. Relying on off-the-shelf AI from big providers can be convenient but locks companies into high ongoing costs. Now, firms like Starbucks are moving to **develop more AI in-house to save on licensing fees** – aiming to trim roughly $10 million annually and reduce dependence on vendors like Microsoft and IBM ([5]). By building tailored AI tools internally, businesses hope to control costs and better target their AI spend toward solutions that directly drive their own metrics.

Why are so many AI programs failing to pay off? A core problem is that **many organizations plunged into AI without aligning it to business processes and goals**. The technology may be powerful, but it can’t deliver value if it’s not effectively embedded in operations. A Boston Consulting Group analysis found that the **dominant determinant of enterprise AI ROI is not the choice of model – it’s how thoroughly companies re-engineer workflows and train people to use the AI** ([1]). In short, simply deploying a cutting-edge algorithm means little if employees don’t adapt their ways of working and the AI isn’t woven into day-to-day decision-making.

Another issue is the lack of preparation and governance. Many companies have treated AI projects as one-off experiments, resulting in what Gartner analysts call an **“accidental factory” of disjointed pilots** with lots of new algorithms but few real systems connecting them ([2]). Such fragmented efforts rarely scale. Worse, they can lead to unexpected consequences – including **runaway costs and risks** – when nobody is minding the store. One **global firm learned this the hard way after racking up a $500 million bill in a single month of unrestricted AI usage** ([3]).

Survey data reflects these oversight gaps. A recent poll of finance executives found **30% of companies had not updated their internal controls in the past year to account for AI systems**, and nearly one-quarter said that if an AI-related mishap occurred, **responsibility for it would be unclear** ([4]). With such weak accountability and cost management, it’s little wonder that many AI initiatives never get past the pilot stage. **The bottom line:** without strong governance and a plan to integrate AI into business workflows, expensive AI experiments will continue to underdeliver.

Amid the hard headlines, there are bright spots and emerging playbooks for success. Experts advise adopting a more disciplined approach to AI investments, starting with financial oversight. For example, **McKinsey is promoting “AI FinOps”** – adapting the financial management techniques used to wrangle cloud computing costs – as a way to track AI expenses and returns more rigorously ([1]). By treating AI projects like a portfolio that requires clear accountability and cost-benefit analysis, companies can avoid waste and focus on initiatives that move the needle.

In addition, new research has identified shared behaviors among companies that are actually realizing value from AI. In a recent Google Cloud survey of 2,400 executives, only **26% of organizations emerged as “AI ROI Leaders,”** meaning they’re seeing accelerating financial returns from AI initiatives ([2]). What sets these high performers apart can serve as a blueprint for others. They consistently do three things to **turn AI into cost-effective growth**:

- Establish **clear ownership and decision-making authority** for AI projects (48% of leading firms have “extremely clear” ownership of AI initiatives) ([3]).

- **Embed AI into core business processes and products** (about half of AI ROI Leaders say AI is embedded in core workflows or enabling new revenue opportunities) ([4]).

- **Mandate ongoing AI training and upskilling** for employees (38% of AI-leading companies provide comprehensive, continuous AI education for staff) ([5]).

By focusing on these fundamentals – governance, integration, and people – a handful of companies are finally translating AI into competitive advantage and measurable returns. For senior leaders, the takeaway is clear: capturing AI’s upside requires pairing technological innovation with operational discipline. Those willing to **align AI projects tightly with business strategy, invest in talent and change management, and rigorously manage ROI** are beginning to see the payoff, while others continue to chase hype with little to show for it.