Why Inefficient AI Spending Today May End Up Driving Tomorrow’s Economic Growth

The world is in the middle of a massive technology investment surge, and artificial intelligence sits at the center of it. Companies across industries are pouring money into AI infrastructure, data centers, software tools, and research teams at an unprecedented pace. Estimates suggest that global AI-related spending could exceed $1.5 trillion in 2025 and climb to nearly $2 trillion by 2026, roughly 2% of global GDP. Yet despite all this spending, the immediate results look underwhelming. Only around 5% of firms currently using AI report clear productivity gains.

To many critics, this disconnect feels familiar. It echoes past investment booms like the dot-com bubble of the late 1990s or the early clean energy rush, where money flooded into new technologies long before sustainable returns appeared. On the surface, today’s AI frenzy looks inefficient, wasteful, and possibly destined for disappointment.

But new research suggests that this apparent inefficiency may actually be a critical driver of long-term economic growth.

Rethinking What “Inefficient” Investment Really Means

In traditional economic thinking, investment efficiency is straightforward. Capital should flow to places where it earns the highest immediate return. If a company invests heavily but fails to generate strong profits or productivity, economists often label that behavior as misallocation of capital.

However, real-world innovation rarely follows such clean rules.

History offers plenty of examples of companies that looked inefficient for years before transforming entire industries. Tesla burned cash for over a decade, Amazon reinvested relentlessly while posting thin or negative profits, and Amgen poured resources into biotech research long before blockbuster drugs emerged. In each case, early investments seemed hard to justify by conventional metrics, yet those investments laid the groundwork for extraordinary long-term success.

This pattern is exactly what a new academic study aims to explain and quantify.

Inside the Study That Challenges Conventional Wisdom

The research, titled Investing in Misallocation, was conducted by Mete Kılıç, assistant professor of finance and business economics, and Şelale Tüzel, professor of finance and business economics. Published in the Journal of Financial Economics, the study analyzes nearly 50 years of firm-level data from most publicly traded U.S. companies.

The researchers identified a surprising pattern. Roughly one in five firms consistently invests heavily even when its current productivity falls below the median. By standard definitions, these firms appear inefficient. Their returns on capital are lower than average, and their investments don’t immediately translate into higher output.

But these firms are not simply reckless spenders.

The Firms That Look Inefficient but Act Like Innovators

When the researchers examined these high-investment, low-productivity firms more closely, several important characteristics stood out.

First, these companies tend to be younger and far more innovation-driven than their peers. They are often operating in emerging industries or developing entirely new products rather than refining established ones.

Second, they are about four times more likely to experience a major performance jump compared to other firms. These jumps are not subtle improvements. Within a few years, such firms often double their sales and increase productivity by around 50%.

Third, these companies are powerhouses of innovation. They file more than twice as many patents as other firms, and those patents receive over three times as many citations, a strong signal that their ideas are influential and widely used. Many of these firms are firmly in the product-innovation phase of their life cycle, where experimentation, trial-and-error, and heavy investment are unavoidable.

In other words, what looks like inefficiency is often intentional risk-taking.

How “Bad” Investments Can Lift the Whole Economy

The study doesn’t just look at individual companies. It also examines what happens at the broader economic level when these firms invest aggressively.

The findings are striking. When high-investment, low-productivity firms increase spending, overall productivity growth across the economy rises over the next five years. These companies may stumble individually, but collectively they act as engines of future growth, pushing technological boundaries and creating spillover benefits for other firms.

Simply put, the economy benefits when some firms are allowed to invest ahead of proven returns.

The Role of Productivity “Jumps”

To explain these outcomes, the researchers built a model centered on the idea of productivity jumps. Not all investments produce steady, predictable returns. Some investments carry a small chance of delivering transformative breakthroughs.

In the data, the average firm has about a 1.6% annual chance of experiencing a major productivity jump. For firms that invest heavily despite low profitability, that probability rises to around 4%, nearly three times higher.

Investing in new technology, research, brand development, or organizational capacity increases the odds of these jumps. While most investments fail to pay off, a few succeed spectacularly, more than compensating for the losses.

When researchers simulated an economy where firms ignored the possibility of these jumps and invested only based on current returns, investment patterns looked cleaner and more balanced. But there was a catch: aggregate productivity declined.

Efficiency, it turns out, can be counterproductive.

Why This Matters for the Current AI Boom

This framework helps make sense of today’s AI investment wave. AI spending is enormous, and the outcomes are uncertain. Many companies experimenting with AI tools, models, and infrastructure will never see direct financial returns. Projects will be abandoned, products will fail, and some investments will simply vanish.

Yet a small number of firms are likely to develop new platforms, tools, and business models that fundamentally reshape industries. From an economy-wide perspective, that is exactly how technological revolutions unfold.

The current divide between companies that are spending aggressively on AI and those taking a wait-and-see approach is not necessarily irrational. According to the study, it may be a natural and necessary part of innovation.

Lessons for Investors and Policymakers

One of the study’s strongest messages is a warning against acting too quickly to “correct” perceived inefficiencies. Policymakers and investors often feel pressure to redirect capital away from firms with low productivity metrics. But doing so could unintentionally stifle experimentation and reduce the chances of future breakthroughs.

The research suggests that some degree of misallocation is not only inevitable but essential. Economies that tolerate short-term inefficiencies may be better positioned to achieve long-term growth.

A Pattern Repeated Across History

Looking back over the past half-century, the U.S. economy has repeatedly followed this pattern. The computer revolution of the 1980s, the internet build-out of the 1990s, and the cloud computing surge of the 2010s all involved waves of heavy investment that initially appeared excessive. Each wave produced failures and waste, but also built the infrastructure and knowledge base that powered future productivity gains.

AI appears to be following the same trajectory.

The central takeaway from the study is clear: an economy obsessed with short-term efficiency risks sacrificing long-term progress. Firms that look unproductive today may be quietly planting the seeds for tomorrow’s breakthroughs.

Research paper: https://doi.org/10.1016/j.jfineco.2025.104208