Skip to main content
EDIT

Why “Inefficient” AI Spending May Power Future Growth

Why “Inefficient” AI Spending May Power Future Growth

New research finds companies investing heavily in new technologies despite low returns are often the ones driving tomorrow’s economic progress.

by: Faculty Perspectives
12.12.25
decorative image of a futuristic graph

[iStock Photo]

Featured

Stay Informed + Stay Connected

MARSHALL MONTHLY BRINGS YOU ESSENTIAL NEWS AND EVENTS FROM FACULTY, STUDENTS, AND ALUMNI.
SIGN UP

The world is in the middle of an investment frenzy. Global spending on AI infrastructure is expected to top $1.5 trillion in 2025 and could reach $2 trillion by 2026 — about 2% of global GDP. Yet only about 5% of firms using AI report clear productivity gains so far.

To many observers, this looks like déjà vu — a replay of the dot-com boom or the early clean-tech rush: too much money chasing unproven ideas. But a new study suggests that what looks like inefficiency on the surface can actually be the engine of long-term progress.

When “bad” Investments are good for growth

In traditional economics, an investment is considered efficient if every dollar of invested capital earns about the same return as it would anywhere else. When a company’s return on capital falls short, it’s often labeled as a misallocation of resources.

But history suggests otherwise. Companies like Tesla, Amgen, and Amazon spent years investing heavily long before their performance justified it. Those early, unprofitable years laid the foundation for the extraordinary growth that followed.

The study, “Investing in Misallocation,” published in the Journal of Financial Economics by Mete Kılıç, assistant professor of finance and business economics, and Şelale Tüzel, professor of finance and business economics, puts data behind this pattern.

Analyzing nearly 50 years of firm-level records from most publicly traded U.S. companies, the researchers found roughly one in five firms invest heavily even when their current productivity is below the median — what standard metrics would call “misallocation.”

These firms behave very differently from the rest. They are typically younger and more innovative, and they are about four times more likely to experience a major leap in performance — doubling their sales and boosting productivity by around 50 percent within a few years.

They also stand out as engines of innovation. These high-investment, low-productivity firms file more than twice as many patents as others, and their patents receive over three times as many citations. They are much more likely to be in the “product-innovation” phase of their life cycle, actively developing new products and technologies.

The broader economic impact is also clear. When these firms increase their investment, overall productivity growth over the next five years rises in a meaningful way. Simply put, those “inefficient” firms are often the ones pushing the economy forward tomorrow.

A new study suggests that what looks like inefficiency on the surface can actually be the engine of long-term progress.

A simple idea with big consequences

To understand this pattern, researchers built a model that showed some companies had a small but meaningful chance of a transformative leap in productivity, a “jump.” Investing in new technology, research, or brand capacity raises the odds of such breakthroughs. In the data, a typical firm has a 1.6% annual probability of a productivity jump, but for the heavy-investing firms with lower profitability, that probability rises to about 4% — nearly three times higher.

When researchers simulated an economy where firms ignore this possibility and invest solely based on current returns, overall investment looks neater and more balanced but aggregate productivity declines.

The data also show that those firms experiencing large productivity jumps tend to peak during major technological transitions — most notably in the late 1990s during the internet boom, and again in recent years as digital and AI technologies have surged. These waves of heavy, risky investment often precede large gains in productivity.

What this means for the AI boom

This framework helps explain today’s AI investment wave. The scale of spending is enormous, and the payoffs are uncertain. Many companies experimenting with AI will never see a direct return. But a few will likely make breakthroughs in the form of new platforms, tools, or business models that transform entire industries. From an economy-wide perspective, that’s how technological revolutions unfold.

The study suggests the current divide — some firms spending aggressively on AI while others hang back — is not necessarily a sign of irrational exuberance. It’s part of the natural process of innovation. When a new technology has the potential to reshape production, broad and uneven investment is not a bug; it’s a feature.

Researchers argue that policymakers and investors should be cautious about declaring inefficiency too soon. Redirecting capital away from seemingly low-productivity firms might inadvertently choke off the experiments that lead to future breakthroughs. The lesson from history, reinforced by the study, is that some degree of “misallocation” is not only inevitable but essential for technological advancement.

The bigger picture

Over the past half-century, U.S. economic growth has often been preceded by waves of seemingly excessive investment: the computer revolution of the 1980s, the internet build-out of the 1990s, and the cloud-computing surge of the 2010s. Each wave produced failures and inefficiencies, but also created the infrastructure and knowledge base for the next era of productivity. AI appears to be following the same pattern.

The study’s takeaway is simple: An efficient economy must tolerate a little inefficiency in the short run. Firms that appear unproductive today may be planting the seeds of tomorrow’s breakthroughs.

RELATED