In a context where artificial intelligence (AI) is at the heart of innovation and growth strategies, business leaders show a surprising astonishment towards the actual results of their investments. Despite a continuous flow of colossal expenses injected into AI, the promise of rapid profitability and spectacular performance improvement struggles to materialize. In 2026, this persistent gap between expectations and reality calls for deep reflection on the effective economic value of these technologies. Far from a simple temporary slowdown, signs of a financial bubble are emerging, marking a paradox between the enthusiasm of leaders and the weakness of profits.
CEOs, while remaining convinced of AI’s potential, express growing unease regarding the absence of tangible returns on investment. A significant portion of them admit to seeing no real positive impact on revenues nor on cost reductions, despite substantial expenditures on AI infrastructures and tools. This dynamic raises major questions: what are the technical, organizational, and strategic barriers preventing AI from becoming a strong lever for corporate economy? And what lessons should be drawn from this disappointment even as the pace of investments does not slow down?
- 1 Massive spending on AI without immediate financial return: a worrying imbalance
- 2 The leaders’ paradox: fearing both the bubble and underinvesting in AI
- 3 Technological and organizational barriers to AI profitability
- 4 Economic impact on a global scale: a shared concern
- 5 Business adaptation: investing in human capital and AI governance
- 6 Perspectives for 2026: between hype and economic reality
- 7 Unknown risks of generative AI on economic performance
- 8 Artificial intelligence and the transformation of leaders’ expectations
Massive spending on AI without immediate financial return: a worrying imbalance
The latest data from a large survey conducted by PwC among 4,454 CEOs illustrates a complex and nuanced picture. More than half of these leaders admit to perceiving no financial return linked to their investments in artificial intelligence. Over the past twelve months, only 30% observed tangible revenue growth attributed to AI, while 56% identified neither increased revenues nor reduced costs.
This situation creates a major gap between the historic level of spending and the economic results recorded. Tens, if not hundreds, of billions of dollars have been injected into building data centers, purchasing specialized hardware, and implementing energy-intensive infrastructures. Yet, revenues from AI remain largely hypothetical, confined to a minority of actors capable of turning this technology into a real economic driver.
The causes of this imbalance
Several factors explain this disconnect between colossal investments and measurable profits. First, many companies advance without a clear roadmap or coherent AI integration strategy. Their projects often remain experimental, limited to prototypes or pilots that do not scale.
Next, organizational and human barriers weigh heavily. Adopting AI requires deep adaptation of business processes, proper employee training, and strengthened data governance. However, to date, 76% of workers have still not received AI training, according to a recent survey.
Finally, technical limitations remain pressing. Generative AI still makes errors, and the complexity of their integration into administrative or decision-making tasks slows profitable deployment. Data security risks also factor in, hindering adoption in certain sensitive sectors.
The leaders’ paradox: fearing both the bubble and underinvesting in AI
A paradoxical tension animates senior leaders: while many fear a bursting of the speculative bubble around AI, they simultaneously worry about not committing enough resources to avoid falling behind technologically.
Mohamed Kande, Global Chairman of PwC, perfectly summarizes this ambivalence: “A limited number of companies are already managing to generate concrete financial returns thanks to AI. But the majority still struggle, which impacts confidence and competitiveness in a global market where the innovation race is intense.”
This fear of irreversible delay pushes many leaders to maintain or even increase their AI budgets, despite profitability that is slow to materialize. This “headlong rush” effect leads to digital indebtedness that could further weaken corporate financial statements.
The AI bubble compared to the 2000s bubble
The historical parallel with the late 1990s Internet bubble has become a recurring theme in current economic analyses. Just like the dotcom bubble, the AI bubble sees massive investments that often precede the emergence of viable and profitable economic models.
However, the size of current investment even exceeds the proportions of that era. While American venture capital invested the equivalent of $344.5 billion between 1997 and 2000, the AI sector had already raised $338.3 billion in 2025, nearly half of which was allocated to generative AI. These colossal sums intensify the risks of a burst, with potentially dramatic economic consequences.
Technological and organizational barriers to AI profitability
One of the major obstacles to transforming AI into a profitability driver lies in the very complexity of its integration into business processes. The technology, although advanced, remains too often poorly adapted to the daily realities of organizations.
A 2025 MIT study revealed that nearly 95% of generative AI projects in companies failed to accelerate revenue. This high failure rate is explained by several technical and human difficulties compounded by poor data management.
Practical example: integrating generative AI into administrative tasks
A multinational bank, despite having invested massively in generative AI to automate its customer service, saw stagnant gains and even increasing costs due to the constant need for manual corrections and human interventions to compensate for system errors. This phenomenon illustrates the difficulty of converting technological innovation into sustainable financial gains.
Furthermore, this case highlights the importance of adequate team training: without sufficient skills to fully exploit these tools, implementation remains ineffective and costly.
The challenges of data management
Reliability and data quality are inseparable from the success of AI projects. Yet many companies struggle to structure and clean their databases, resulting in biased or unusable outcomes. The deficiency in this area represents a major handicap.
Consequently, several organizations prefer to prioritize safety and security, at the expense of effective AI exploitation, especially in regulated sectors where data protection is paramount.
The issue is not limited to companies alone. This trend also has repercussions on the global economy. The risk of a technological bubble bursting could affect financial markets, investment funds, and more broadly, confidence in technological innovation.
Banks and venture capitalists observe both the growing opportunities and risks that AI represents. The volatility of valuations and stagnant profitability require a reassessment of financing and diversification strategies.
Table: Comparison of investments and returns between dotcom bubble and AI bubble
| Criterion | Internet Bubble (1997-2000) | AI Bubble (2023-2025) |
|---|---|---|
| Amount invested (in billions of dollars) | 344.5 | 338.3 |
| Proportion dedicated to main technology | 100% Web | Nearly 50% Generative AI |
| Investment duration (years) | 4 | 3 |
| Measurable short-term financial return | Low, especially in the last 2 years | Mostly low, exception for few leaders |
| Impact on the global economy | Formation of financial bubbles | Increased risk of bubble burst with contagion effect |
Business adaptation: investing in human capital and AI governance
Faced with these challenges, organizations have understood that the technological dimension alone is not enough to guarantee profitability. Integrating AI effectively requires a holistic approach, focused on people and processes. Investments in training, organizational adaptation, and data governance become crucial.
Training as an essential lever
An alarming 76% of employees report not having received AI-related training. This deficit severely hampers the appropriation of tools and the ability to derive a competitive advantage from them.
Customized training programs, practical workshops, and enhanced awareness are some of the solutions already implemented by pioneering companies. These initiatives promote better understanding of uses and greater confidence in the technology.
Strengthening data and AI project governance
Data quality and management, as well as project governance, are among the priorities to improve AI-related performance. Ethical frameworks, clear accountability rules, and rigorous controls are deployed to ensure compliance and secure projects.
These measures also help optimize processes and reduce hidden costs linked to errors and misuse of systems.
Perspectives for 2026: between hype and economic reality
While enthusiasm around AI remains intact among decision-makers, efforts focus on turning promises into concrete results. 2026 will be a pivotal year where pressure on profitability and performance will intensify. Companies will need to demonstrate that their investments are not merely a heavy burden but a solid value creation driver.
Success likely lies in the synergy between technological innovation, team training, organizational adaptation, and rigorous project management. Only organizations capable of mastering these levers can claim sustainable positive financial returns.
Ways to maximize ROI in AI
- Develop a clear strategy with measurable objectives and a precise roadmap.
- Strengthen training of employees for effective AI tool integration.
- Optimize data governance and ensure their quality and security.
- Foster an agile innovation culture to speed up implementation and adaptation based on feedback.
- Regularly measure results to adjust strategies in real time.
Unknown risks of generative AI on economic performance
As generative AI concentrates a large share of investments, it also presents insufficiently anticipated potential dangers. Frequent errors, algorithmic biases, and confidentiality issues can harm overall efficiency and reduce expected profitability.
Poor use of these technologies can generate significant hidden costs related to error correction, legal disputes, or loss of trust from clients and partners.
The five major dangers to watch
- Persistent errors generated by imperfect or unsuitable models.
- Algorithmic biases that can reproduce or amplify discrimination.
- Confidentiality problems exposing organizations to risks of data leaks.
- Excessive dependence weakening critical human skills.
- Lack of transparency making it difficult to identify causes of errors or malfunctions.
Artificial intelligence and the transformation of leaders’ expectations
As technology evolves, leaders’ expectations are being redefined. The initial surprise at the lack of profits has turned into a clearer will to focus efforts on sustainable profitability and shared value creation.
Companies now seek to deploy AI not as a miraculous panacea, but as a complementary tool integrated into a global strategy oriented towards efficiency, security, and service quality.
Keys to adapted AI performance
Among the clear identified factors, the ability to leverage AI to improve customer experience ranks high. Only 10% of companies today manage to truly use AI to transform customer interaction and produce a commercial impact.
This customer-centric approach encourages leaders to rethink investments by favoring tailored solutions and rigorous results management.
Why doesn’t AI generate the expected profits yet?
AI does not yet produce the expected profits mainly due to a lack of precise strategy, technological integration issues, personnel training deficits, and insufficient data governance. Many projects remain at the pilot stage without scaling to profitable exploitation.
How do leaders perceive AI investments despite limited returns?
Despite often disappointing financial returns, leaders maintain or increase their investments for fear of falling behind competitors, creating a paradoxical dynamic between caution and a desire to innovate.
What are the main technical barriers slowing AI profitability?
Technical limits include frequent errors, algorithmic biases, difficulty integrating into existing processes, and data quality and security problems that slow deployment of profitable solutions.
What measures can help improve AI economic performance?
To maximize return on investment, it is crucial to develop a clear strategy, train staff, optimize data governance, establish an agile innovation culture, and regularly measure results to adjust actions.