Shifting Investment Benchmarks Amid AI Bubble Fears, Profitability Takes Precedence Over Growth

PEF Investment Strategies Pivot Toward Profitability in AI
Preference for Tangible Assets with Stable Demand Intensifies
Big Tech Debt Expansion Fuels Rising Concerns Over AI Bubble

As the artificial intelligence (AI) investment boom persists, capital continues to concentrate in infrastructure sectors such as semiconductors and data centers. However, private equity (PEF) firms are adopting a markedly different approach. Rather than targeting AI technology itself, they are seeking investment opportunities in assets where AI adoption enhances revenue structures. With overheated valuations converging with rapidly expanding leverage, concerns over a potential market-wide bubble are intensifying. This shift is interpreted as a strategic effort to mitigate potential capital losses and secure financial stability.

Private Equity Firms Betting on AI ‘Revenue Models’

According to the investment banking (IB) industry on April 16, global PEF firms are increasingly prioritizing monetization potential over mere growth prospects when investing in generative AI companies. Blackstone, for instance, invested approximately $200 million in generative AI startup Anthropic last month. This brings Blackstone’s cumulative investment in the company to $1 billion, with Anthropic now valued at around $350 billion.

Anthropic is rapidly advancing commercialization across various domains, including workflow automation and development support, based on its enterprise AI model “Claude.” In particular, its subscription-based model targeting corporate clients has enabled steady revenue expansion, establishing a monetization framework grounded in enterprise demand.

Permira is another notable player demonstrating active investment strategies. Clearwater, a cloud-based financial software company backed by Permira, serves as a core system for processing large-scale financial data and handling regulatory reporting. The company benefits from a strong lock-in structure that makes customer attrition difficult. As data accumulation directly enhances performance, increased AI adoption simultaneously strengthens both service value and profitability.

Similarly, Permira’s portfolio company, education administration platform Ducky, presents a comparable case. Operating in environments where clients such as schools face challenges in building proprietary systems, Ducky secures stable demand. Its per-student pricing model enables revenue expansion alongside the rollout of AI-driven functionalities. Infrastructure investments follow a similar trajectory. Assets such as Permira’s data center platform, Fleet, directly benefit from rising AI demand while generating stable cash flows through long-term contracts.

Capital Shifts Toward Businesses Resistant to AI Disruption

Within parts of the PEF industry, a notable capital shift is emerging—from traditional software (SaaS) investments toward industrials and infrastructure-based tangible assets. This trend is referred to as the “HALO (Heavy Asset, Low Obsolescence)” trade. The concept emphasizes assets that require substantial capital investment but face relatively low risk of rapid obsolescence due to technological change. It is gaining traction across the broader global investment landscape.

HALO assets, by definition, are “capital-intensive yet durable.” Representative categories include industrials, energy, power grids, data centers, and minerals—asset classes grounded in physical infrastructure. While their growth potential may be more limited compared to software, they offer lower substitution risk and more predictable long-term demand. According to major international media reports, these characteristics are increasingly being recognized as premium factors in the current market environment. In fact, major global PEF firms such as Bain Capital, Brookfield, and Blackstone are reassessing their approach to software investments, while large-scale transactions involving industrial assets continue across European markets.

In contrast, the software investment market is showing clear signs of deceleration. Delays and cancellations in corporate sales and initial public offerings (IPOs) are increasing uncertainty in the exit market, while also amplifying valuation pressures on existing holdings. PEF firms that significantly expanded their software exposure in recent years are now facing growing pressure to reassess their portfolios.

These dynamics extend beyond equity investments. The leveraged buyout (LBO) structures built around software companies are beginning to show signs of strain, spilling over into credit markets. Given the typically high leverage in software-related lending, heightened valuation uncertainty is rapidly reinforcing risk aversion among investors. In this process, the criteria for investment decision-making are also evolving. Market participants note that the current shift is particularly significant in that it reflects a fundamental reassessment of what merits a valuation premium. Increasingly, global investors are placing greater value on “tangible assets” and “clear demand foundations” over high-growth narratives.

AI Investment Frenzy and Escalating Bubble Concerns

The primary reason PEF firms are taking a more cautious stance toward AI-related investments lies in growing concerns over a potential AI bubble. Traditionally, big tech companies relied heavily on internal free cash flow for funding. Recently, however, the large-scale deployment of data centers for AI development has driven a transition toward unprecedented levels of external debt financing.

According to the Organisation for Economic Co-operation and Development (OECD)’s Global Debt Report, total borrowing by governments and corporations worldwide is projected to reach $29 trillion this year, a 17% increase compared to 2024. Sovereign bond issuance by OECD member central governments alone is estimated at $18 trillion, with 78% of that amount intended for refinancing existing debt.

A key driver of this trend is the transformation of U.S. hyperscalers—once known for their massive cash-generating capabilities—into large-scale borrowers. The OECD reported that nine major hyperscalers, including Alphabet, Amazon, Apple, Microsoft, Meta, Tencent, Alibaba, IBM, and Oracle, collectively raised $122 billion in the global bond market last year. This figure accounts for 45% of total bond issuance by all technology companies worldwide.

Big tech firms are not only increasing bond issuance but are also employing increasingly complex financing structures. A representative example is Meta’s “Hyperion” hyperscale data center project in Louisiana, with a total investment of $30 billion. Meta’s funding strategy exemplifies a “Frankenstein-style” approach, combining private equity financing, corporate bonds, and project financing (PF)—a method traditionally used in real estate. The complexity reflects Meta’s already substantial debt load; the company had previously raised $30 billion through corporate bonds in October last year for operational purposes.

Such massive capital expenditures would not pose a significant issue if they did not strain cash flows. However, these companies are already experiencing stagnation in free cash flow. Without a corresponding surge in revenue, continued capital spending risks becoming unsustainable. For these investments to be justified, future AI-related revenues must increase dramatically. Bain Capital estimates that big tech firms would need to generate annual revenues of $2 trillion by 2030 to justify current data center investment levels. More than three years after the launch of ChatGPT, AI-related revenues for these companies are estimated at $20 billion—implying that revenues would need to increase more than 100-fold to recoup investment costs.

If supply begins to outpace demand, companies will be forced to absorb the aftereffects of massive overinvestment in production capacity. Valuations could contract sharply, while firms that issued debt to finance capital expenditures may face insolvency risks. The current AI investment boom can be interpreted within this broader context. While there is little disagreement over the long-term growth trajectory of AI demand, experts broadly agree that there is no guarantee this demand will translate into actual revenue and profit generation.