The corporate world has spent the last three years in a frenzy of artificial intelligence investment. Billions poured into infrastructure, talent, and licensing deals. Boardrooms echoed with promises of transformation. And now the bill is coming due — with precious little to show for it.
According to a new forecast from Gartner, at least 30% of generative AI projects will be abandoned after the proof-of-concept stage by the end of 2025. The reasons range from poor data quality to inadequate risk controls to murky business value. That’s not a rounding error. That’s nearly a third of corporate AI bets failing before they even reach production, as The Register reported.
The numbers get worse from there. Gartner’s analysis suggests that most enterprises won’t achieve meaningful financial returns from their generative AI deployments until 2028 at the earliest. For an industry that has been selling AI as an imminent productivity miracle, three more years of waiting is a bitter pill.
Rita Sallam, a distinguished VP analyst at Gartner, didn’t mince words. She noted that many organizations jumped into AI initiatives without clearly defining what success would look like. The result: sprawling pilot programs with no measurable impact on revenue or cost structures. “The gap between AI ambition and AI value is widening, not narrowing,” Sallam said during Gartner’s recent IT symposium, according to The Register.
This isn’t the first time enterprise technology has followed this pattern. Cloud computing went through a similar cycle of inflated expectations followed by painful rationalization. But the AI spending spree has been faster, bigger, and more heavily marketed than anything that came before it. Companies that might have spent months evaluating a cloud migration spent weeks greenlighting AI deployments — often under pressure from CEOs who’d read one too many breathless magazine covers.
The scale of capital commitment is staggering. Microsoft, Google, Amazon, and Meta have collectively pledged more than $200 billion in AI-related capital expenditure for 2025 alone. That figure, reported widely across financial media, represents a bet that enterprise demand for AI compute will continue to accelerate. But Gartner’s findings raise an uncomfortable question: what if demand stalls because customers can’t figure out how to make AI pay for itself?
Some early warning signs are already visible. A January 2025 survey by Boston Consulting Group found that roughly half of C-suite executives were “dissatisfied” with their AI investments to date. Not disappointed. Dissatisfied. The distinction matters. Disappointment implies patience. Dissatisfaction implies budget cuts are coming.
And budget cuts may already be underway in pockets of the market. Several enterprise software companies have quietly scaled back AI-specific sales teams in recent months. Others have shifted their messaging from “AI-first” to “AI-enhanced,” a subtle but telling retreat from the maximalist positioning that dominated 2023 and 2024.
The technical barriers are real. Generative AI models require clean, well-structured data to deliver reliable outputs. Most enterprises don’t have that. They have decades of legacy systems, inconsistent data formats, and organizational silos that make integration painful and expensive. Gartner’s research highlights data quality as the single biggest obstacle to AI deployment success, a finding that will surprise exactly no one who has worked inside a large corporation’s IT department.
Then there’s the hallucination problem. Large language models still fabricate information with alarming confidence. For consumer applications like chatbots and writing assistants, occasional errors are tolerable. For enterprise use cases — legal analysis, medical records, financial reporting — they’re potentially catastrophic. Companies that deployed AI without adequate guardrails are now scrambling to add them retroactively, which is both more expensive and less effective than building them in from the start.
Risk and compliance concerns compound the challenge. Regulatory frameworks for AI remain fragmented and fast-moving. The European Union’s AI Act is beginning to take effect in stages. The United States still lacks comprehensive federal AI legislation, though several states have introduced their own rules. For multinational corporations, this patchwork creates a compliance headache that slows deployment timelines and increases legal costs.
Not every company is struggling, of course. Organizations with strong data infrastructure and clear use cases — think fraud detection in banking, predictive maintenance in manufacturing, or drug discovery in pharmaceuticals — are seeing genuine returns. But these tend to be companies that were already sophisticated users of machine learning before the generative AI wave hit. They had the plumbing in place. Most didn’t.
The consulting industry deserves some scrutiny here. Firms like McKinsey, Deloitte, and Accenture aggressively marketed AI transformation services beginning in late 2022, often projecting enormous productivity gains based on theoretical models rather than empirical results. A widely cited McKinsey report from June 2023 estimated that generative AI could add $2.6 trillion to $4.4 trillion in annual value to the global economy. Those numbers helped justify enormous corporate spending. Whether they’ll prove accurate is another matter entirely.
Gartner’s more sober assessment reflects a growing consensus among analysts that the AI hype cycle has peaked — at least for this iteration. The research firm’s famous Hype Cycle model would place generative AI somewhere in the “Trough of Disillusionment,” the painful period after initial excitement fades and before practical, sustainable adoption takes hold. Getting through that trough typically takes two to five years. Which lines up neatly with Gartner’s 2028 timeline for meaningful returns.
Wall Street is starting to pay attention. Investors who bid up AI-adjacent stocks to extraordinary valuations in 2023 and 2024 have grown more selective. Shares of several enterprise AI startups that went public via SPAC have lost 60% or more of their value. Even the hyperscalers — Microsoft, Google, Amazon — have faced pointed questions from analysts about when their massive AI capital expenditures will translate into proportional revenue growth.
Satya Nadella addressed this directly on Microsoft’s most recent earnings call, arguing that AI monetization follows an “S-curve” pattern and that the company is still in the early, steep part of the investment phase. That may be true. But patience is a finite resource on Wall Street, and the clock is ticking.
One area where returns have materialized faster than expected is developer productivity. GitHub Copilot, Microsoft’s AI-powered coding assistant, has attracted more than 1.8 million paying subscribers as of early 2025 and demonstrably speeds up certain coding tasks. But developer tools are a relatively narrow slice of the enterprise AI market. The bigger prize — automating knowledge work at scale across functions like sales, marketing, HR, and finance — remains elusive for most organizations.
There’s also a talent problem. Companies need people who understand both the technical capabilities of AI and the specific business processes they’re trying to improve. Those people are rare and expensive. Many organizations have hired data scientists and machine learning engineers only to discover that the bottleneck isn’t model building — it’s change management, process redesign, and executive alignment. Technical talent without organizational authority produces impressive demos and not much else.
So where does this leave the industry? In a period of recalibration. The companies that will emerge strongest are those willing to do the unglamorous work: cleaning their data, defining clear metrics for success, starting with narrow use cases, and scaling only what works. The ones most at risk are those still chasing the hype, throwing money at AI initiatives because their competitors are, without a clear theory of how those investments will generate returns.
Gartner’s 30% abandonment rate, as reported by The Register, should serve as a wake-up call. Not because AI lacks potential — it clearly has enormous potential — but because potential without execution is just an expense line. And for a growing number of CFOs, that expense line is getting harder to justify.
The AI boom isn’t over. But the easy-money phase is. What comes next will be slower, messier, and more selective. Companies that treat AI as a long-term capability investment rather than a quick fix will eventually see returns. The rest will join the 30% — writing off millions in sunk costs and wondering what went wrong.
History suggests the technology will ultimately deliver. But history also suggests that the path from hype to value is longer, more expensive, and more littered with failures than anyone wants to admit while the hype is still running hot.
WebProNews is an iEntry Publication