Big AI bets, little returns: the ROI gap hits home.

Companies in 2026 are investing in artificial intelligence at an unprecedented scale, but gains remain stubbornly hard to find. Worldwide AI spending is forecast to reach about $2.59 trillion this year – a staggering 47% jump over 2025’s level ([1]). Yet, according to a major global survey, only 37% of organizations have seen even a modest positive impact on their earnings from AI initiatives so far, essentially the same share as last year ([2]). Even more striking, the proportion of true standouts – those where AI drives more than 5% of company EBIT – is stuck at roughly 6% of firms ([3]). In short, enormous investments in AI have not translated into meaningful bottom-line returns for the vast majority of businesses.

This disconnect is underscored by the contrast between pervasive adoption and lackluster results. In McKinsey’s 2026 global AI study, nearly nine in ten surveyed companies reported using AI in at least one business function ([4]), and roughly 80% of workers said these tools have improved their personal productivity ([5]). However, those individual efficiency gains often don’t register in the profit-and-loss statement ([6]). Many employees may be completing tasks faster or generating more content with AI’s help, but if those improvements don’t translate into higher revenue, cost savings, or other clear business KPIs, the ROI remains essentially invisible. The result is a growing frustration: even as staff and teams enthusiastically embrace AI tools, senior leaders are asking why all that activity isn’t showing up in the financials.

Industry analysts have a name for this phenomenon: the "AI accountability gap" – the widening chasm between what organizations spend on AI and what they actually get back in measurable value ([7]). This gap represents all the AI-driven effort that can’t be tied to any concrete business outcome, and by all accounts it is large and growing. That realization is now front and center for executives. After years of breathless hype about artificial intelligence, 2026 is shaping up to be the year leaders face a pivotal question: Are these AI investments truly paying off?

The era of blank-check AI projects is coming to an end as financial oversight catches up with the AI hype. During the initial boom, many organizations allowed technology teams to pursue artificial intelligence without the usual rigorous scrutiny – in part due to fear of missing out if competitors leapt ahead ([1]). Earlier this year, a Boston Consulting Group survey found 94% of companies planned to keep boosting AI spending even when immediate returns were uncertain ([2]), reflecting how strong the “must not fall behind” mentality was. But today, that freewheeling approach is giving way to tougher questions. CFOs and boards are reasserting discipline and asking: What are we getting for this money?

In the past 48 hours, evidence of this shift has emerged. Forrester’s latest analysis indicates that roughly one quarter of planned enterprise AI investment is now being postponed until 2027 amid heightened ROI scrutiny ([3]). And in a recent Gartner survey of IT leaders, fewer than one in three could point to any specific financial benefits from their AI implementations so far ([4]). These revelations drive home why many organizations are re-evaluating their AI portfolios. After several years of experimentation, patience is wearing thin. If an AI initiative can’t clearly demonstrate business value – increased revenue, cost savings, or measurable efficiency gains – it may no longer survive the next budgeting cycle.

Even tech trailblazers have felt this reality check. A few months ago, Uber – widely seen as an AI-forward company – provided a cautionary tale. The ride-hailing giant’s COO publicly admitted that ballooning AI costs had become 'harder to justify' than expected ([5]). Uber had reportedly blown through its entire 2026 budget for AI coding tools by April, as 95% of its engineers embraced AI assistance and 70% of new code was generated by machine – yet executives couldn’t tie this to any tangible improvements in customer experience or revenue growth ([6]). That stark example has not been lost on other corporations. It underscores that even aggressive AI adoption does not guarantee payoff, and it has emboldened finance chiefs elsewhere to demand proof of ROI before signing off on further AI spending. The takeaway for leaders: moving forward, every AI dollar will be expected to show a return, or risk getting cut.

The economics of AI are shifting under pressure as organizations hunt for better returns. In the past day alone, two major AI vendors announced steep price cuts for their latest models – effectively kicking off a price war and acknowledging customers’ cost concerns. OpenAI unveiled its new GPT-6 models, nicknamed 'Sol' and 'Luna', at roughly half the per-token cost of its previous GPT-5.6 generation ([1]). And Anthropic launched its updated Claude 5.5 at about 40% less cost than the prior version while claiming comparable performance gains ([2]). These aggressive cuts aim to lower barriers for enterprise adoption and suggest that even leading AI providers feel the need to help clients improve the ROI of AI by reducing the “I” (investment) side of the equation.

However, lower price tags on model usage don’t necessarily solve the cost challenge. A recent industry analysis reveals that AI services are unlikely to see significant underlying cost relief until new hardware catches up – likely not until 2027 ([3]). Next-generation AI chips and GPUs that promise better efficiency won’t be widely deployed in data centers for another 18 months or more, meaning cloud providers currently face little hardware-driven pressure to drop their prices ([4]). Meanwhile, model complexity and demand keep climbing, pushing the cost of AI inference upward for users even as providers cut rates on paper ([5]). In other words, unless organizations manage their usage, they could still end up with hefty AI bills despite nominally cheaper models. Leaders must be wary of assuming that their AI costs will naturally decline; as one analysis warns, baking in expectations of imminent cost drops is a "structurally flawed" gamble that could undermine business models before savings materialize ([6]).

All this puts a premium on cost discipline and strategic decision-making. For some companies, the response is to explore alternatives – from fine-tuning open-source models in-house to using smaller, task-specific AI solutions – in order to sidestep the highest costs of big proprietary AI platforms. Others are doubling down on vendor negotiations and cloud cost management. Notably, some enterprises are adopting specialized AI FinOps tools and securing multi-year cloud contracts now to lock in better rates, well before 2027’s expected hardware-driven price improvements kick in ([7]). The implication is clear for executives: controlling the costs of AI (through smarter use of resources, partnerships, and procurement) is fast becoming as important to the business case as the promise of AI-driven innovation.

Faced with these pressures, businesses are seeking ways to turn AI experiments into real results. The key is shifting from technology for technology’s sake to a focus on measurable business outcomes. Simply rolling out AI across the organization isn’t a guarantee of ROI – a lesson even early adopters like Uber, Microsoft, and Duolingo have implied in public. All three have reported surging internal AI use, only to question whether all those extra chatbot conversations and auto-generated code are actually moving the needle on revenue or efficiency ([1]). High usage in itself is no panacea; what matters is how that usage is directed toward genuine productivity and profit improvements.

With that in mind, leading organizations are embracing a more disciplined, value-first playbook for AI. One pillar of success is to pre-define metrics and a baseline for every AI initiative, then rigorously track progress. As a recent industry report emphasized, a useful AI ROI review should clearly articulate three things: what was spent, what changed in the business, and how much of that change can be attributed to the AI solution ([2]). Equally important is counting all the costs – including often-overlooked integration work, human oversight, governance, and ongoing cloud fees – to ensure the ROI calculation is realistic ([3]). By demanding this accountability, leadership can catch underperforming projects early and course-correct before sinking more resources.

Another differentiator for the small group of companies achieving strong AI returns is aligning AI with core operations and strategic goals. Analysis of successful deployments shows that organizations realize the highest ROI when they apply AI to fundamental business processes and pair it with process re-engineering, solid data foundations, and active executive sponsorship ([4]). Rather than scattering pilots in silos, these firms scale AI deliberately and manage it with robust oversight. They also invest in building trust and transparency around AI. Notably, new research suggests that companies with mature "responsible AI" practices and governance are far more likely to see substantial returns on their AI investments ([5]). By treating AI as a strategic tool – not a magic wand – and weaving it into the fabric of their business with clear accountability, these leaders are starting to close the ROI gap and achieve tangible performance gains.