"Technology works, but value hasn't arrived": Bain report reveals returns are scant after AI spending surpasses one trillion dollars

As more than $1 trillion in global capital feverishly flows into the artificial intelligence (AI) sector, market attention is inevitably returning to the most core indicator of business logic: Return on Investment (ROI).

The latest global executive survey by Bain, a world-leading consulting firm, pours cold water on the heated wave of AI investment: After collective enterprise AI spending has exceeded $1 trillion, actual cost savings from automation are generally far below expectations.

The survey covers 951 companies worldwide with annual revenues over $100 million, spanning nine major industries: retail, technology, advanced manufacturing, healthcare, consumer goods, energy, financial services, telecom/media/entertainment, and insurance. Results show that among companies able to quantify AI cost savings, the largest group (40%) achieved cost reductions of only 10% or less—falling far short of the grand initial expectations.

Even more alarming: 44% of large enterprises are using “unrealized AI savings from the last round” as rationale for funding the next round of AI investments—Bain terms this “a structurally flawed cycle bet.” Meanwhile, Gartner’s concurrent report forecasts that over 40% of Agentic AI projects will be shut down by the end of 2027. Bain nails the current situation: “The technology is in place, but the value hasn't arrived.”

These signals mean: The valuation logic of AI concept stocks faces a severe challenge—current high valuations are based on “projected value” rather than “actual value.” Once the market seriously calculates ROI (return on investment), the risk of valuation restructuring cannot be ignored.

40% of Companies’ AI Cost Savings Are Less Than 10%—A Serious Gap Between Expectation and Reality

The Bain survey completed this April reveals a reality that should make executives “uneasy”: Among companies able to measure AI cost savings, the largest group (40%) achieved cost reductions of only 10% or less. Yet the investment these companies made in AI technology often far exceeds this level of savings.

Bain states bluntly in its report, these results “should make executives uneasy”—since many of them authorize continually increased AI budgets based on “expected savings.” The core issue: The visible savings effect is negligible.

This isn’t the first evidence that AI underperforms expectations. Last year, MIT released a study showing 95% of corporate AI pilot projects ended in failure, mainly because “tools couldn’t learn, integration was poor, or didn’t match actual workflows.”

“Cycle Bet”: Using Unrealized Savings to Endorse the Next Round of Investment

The most concerning Bain finding is a fundamental flaw in corporate AI investment decision logic: 44% of large enterprises are using “savings” from the previous round of AI investment as the funding rationale for the next round—even though those savings have not been realized.

Bain issues a clear warning: “The savings pool available for allocation is far smaller than expected because the last round failed to deliver on its promise. The commercial reasoning for this round of investment is based on predicted rather than actual value.”

The report further points out that this “self-rolling financing of the next round from past returns” appears to be strict financial discipline on the surface, but “in essence, it is a structurally flawed cycle bet.” Bain’s concise yet striking conclusion: “The technology is in place, but the value hasn't arrived.”

Data Predicament: Hundreds of Billions Spent on Data Modernization Still Can’t Solve AI’s “Data Shortage”

Bain’s report also highlights the primary cause of poor AI project performance—and unexpectedly, it’s a basic problem: Companies can’t reliably access their own data.

“Despite a decade of global investment in data modernization totaling hundreds of billions, the primary cause of poor AI project performance remains that companies can’t steadily access their own data,” Bain writes in the report.

A notable (counterintuitive) finding of the survey: Companies that actually achieved savings goals encounter more obstacles in data structure and accessibility than those that didn’t—but the former report fewer challenges at the organizational level, such as insufficient budgets or conflicting priorities.

Bain’s prescription is: Companies should not wait until all data is organized before feeding AI models, but should start with available data and then use AI to help structure the rest.

Gartner Warning: Over 40% of Agentic AI Projects Will Die Out Before 2027

Echoing Bain’s report is Gartner’s concurrent study. Gartner predicts that over 40% of Agentic AI projects will be cancelled by the end of 2027, for reasons including rising costs, unclear commercial value, and lack of risk control mechanisms.

Anushree Verma, Gartner’s Senior Director Analyst, points out: “Most Agentic AI projects are still in early experiments or proof-of-concept stages, largely driven by market hype and often misapplied. This causes companies to overlook the real cost and complexity of deploying AI agents at scale, leading to project stagnation at production deployment.”

Gartner advises that applications for Agentic AI should be strictly limited to scenarios where they bring clear value or measurable ROI and emphasizes that integrating AI agents into legacy systems is technically very complex, often disrupting existing workflows and resulting in high transformation costs. Verma further notes: “To realize true value from Agentic AI, companies must focus on enterprise-level productivity gains—not just auxiliary enhancements at the individual task level.”

Bain concludes with a statement that embarrasses the entire industry: In this technological wave that has absorbed more than $1 trillion in capital, almost no company has done effective ROI analysis.

“Companies that validate their reinvestment logic using actual—not expected—automation returns are managing risk. Those that don’t are just piling on risk,” Bain warns in its report.

Current high AI-related company valuations are largely based on optimistic forecasts about future earnings, rather than verified actual performance—mirroring the same corporate AI investment logic Bain criticizes: Valuation pricing relies on ‘predicted value,’ not ‘actual value.’

With token costs soaring, companies are sobering up and starting to retreat from the “huge promises” of Agentic AI.

This also explains why AI giants like OpenAI and Anthropic, who permanently extrapolate short-term revenue explosions, are rushing to complete their IPOs before the market re-examines and recalculates true ROI.

Risk Disclosure and DisclaimerThe market carries risks and investments must be made cautiously. This article does not constitute individual investment advice, nor does it consider the specific investment objectives, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article fit their particular circumstances. Investment based on this article is at your own risk.