AI investments under pressure: from hype to hard ROI.

Behind this slow ROI is an economic reality check: deploying advanced AI is extremely expensive. A new IBM study warns of soaring compute bills that could rise 89% from 2023 to 2025, with 70% of executives blaming generative AI's unprecedented demands ([1]). In fact, every executive surveyed reported canceling or delaying at least one AI initiative due to cost concerns ([2]). Even AI's pioneers are feeling the strain – OpenAI itself raised over $6 billion in 2024 at a $157 billion valuation just to fund its own skyrocketing compute needs ([3]).

These ballooning costs are forcing hard choices. Some companies are reevaluating 'build vs buy' decisions as they realize renting AI models and cloud infrastructure can become unsustainably pricey. Others are turning to hybrid cloud setups and technical fixes – from optimizing code to using smaller, task-specific models – to rein in expenses ([4]) ([5]). The bottom line: without careful cost management, AI projects risk becoming financial sinkholes that eat up any efficiency gains they might create.

There's also growing recognition that traditional cost-cutting as an AI 'benefit' may be a mirage. Many organizations leapt to trim headcount, expecting AI automation to immediately save money. Yet a Gartner analysis of large firms deploying AI found no link between workforce reduction and higher returns ([6]). Companies that laid off staff at similar rates to high-ROI peers often saw no added value ([7]). In other words, cutting jobs too early can undermine rather than improve the business case for AI.

The enormous hype around AI is driving sky-high expectations – and anxiety – in boardrooms and stock markets alike. One striking example is startup Anthropic reportedly seeking funding at a jaw-dropping $900 billion valuation ([1]), despite having yet to generate commensurate revenue. This kind of speculative fervor has even led struggling companies to reinvent themselves with an 'AI' label to ride the wave. Allbirds, a shoe retailer, rebranded as 'NewBird AI' and saw its stock explode 582% in days before plummeting back to earth ([2]). Now, seasoned short-sellers are actively hunting for these kinds of 'fake AI' plays, betting that many exaggerated claims will eventually unravel ([3]).

Meanwhile, corporate leaders face intensifying pressure from investors and directors to justify costly AI bets. In one global survey, 53% of investors said they expect to see positive ROI from AI initiatives in six months or less ([4]) – an extraordinarily short timeframe given that meaningful financial impact from AI typically requires 18-36 months, or even up to five years for enterprise-level transformations ([5]). Another study found 71% of CIOs fear their boards will freeze or cut AI funding if results don't materialize within two years ([6]). This urgency from the top is compelling some firms to rush deployments or pursue high-profile projects to 'signal' progress, even if the business case is tenuous.

The result is a growing disconnect between AI rhetoric and reality. Even as CEOs tout AI on earnings calls, many privately acknowledge that direct revenue or cost savings from AI remain limited. When initiatives fail to meet inflated expectations, it not only disappoints stakeholders but can breed internal cynicism, making it harder to fund truly valuable long-term AI capabilities. The current climate rewards bold AI narratives – as seen when Cisco's stock jumped 17% after announcing an AI-focused restructuring alongside strong earnings ([7]) – but the true test will be delivering sustainable business outcomes to back up those narratives.

With so much at stake, a critical question emerges: what are the few organizations getting right with AI? Studies indicate that those capturing outsized returns treat AI not as a magic bolt-on, but as a strategic transformation catalyst for their business. PwC's new AI Performance study found 74% of AI's economic value is concentrated in the top 20% of companies, which use AI to drive new growth and reinvent their business models rather than just cutting costs ([1]). These leaders are also far more likely to redesign core workflows for AI and to automate decision-making within guardrails – while investing in strong data governance to build trust at scale ([2]).

([3])Experts stress that achieving ROI requires rethinking processes and empowering people, not removing them. Returns are strongest when AI is embedded in well-defined operations and paired with process re-engineering and upskilling, rather than scattered across siloed pilot projects ([4]). For example, New York Life's AI program began with a clear directive from the CEO to focus on use cases that directly impact the company's earnings plan, and to reinvest early wins into larger initiatives ([5]) ([6]). By targeting areas with rich data and leadership support – and maintaining financial discipline – they avoided the pilot trap and steadily unlocked value.

Even AI vendors now recognize that integration and human factors are pivotal to success. OpenAI's decision to launch a $4 billion deployment company – effectively an in-house AI integration arm – to embed its engineers into client operations highlights the need to emphasize implementation, change management, and tangible results ([7]). As Gartner's Helen Poitevin concluded, companies that improve ROI are not those that eliminate people, but those that amplify them ([8]) – using AI to boost human productivity and create new capabilities. The message for executives: realize AI's potential by pairing the technology with the right strategy and talent, and measure success by real business outcomes, not buzzwords.