When "pension money" meets "AI fever": What are the risks of global pension funds heavily betting on data center infrastructure?

With the wave of artificial intelligence (AI) sweeping the global financial markets, pension funds and insurance capital, which have long pursued stable returns, are accelerating their entry into AI infrastructure fields such as data centers, attempting to get a share of this emerging technology wave. However, this has raised deep concerns in the market about the safety of public savings funds.

Executives from SoftBank Group recently stated that its "Stargate" data center project in the United States, which is as large as $500 billion, may have potential supporters including life insurance companies and pension funds as long-term investors. SoftBank founder Masayoshi Son is extremely optimistic about this, believing that given AI's significant contribution to the global economy in the future, even a massive investment can quickly recoup costs.

At the same time, the Canada Pension Plan Investment Board also announced a partnership plan with Australia's Goodman Group, with both parties set to invest billions of dollars in data center projects in Europe. Analysts believe that this series of actions indicates that, despite the uncertainty of future profit models for AI technology, tech giants in urgent need of funding to build expensive computing power facilities are aligning perfectly with institutional funds seeking long-term asset allocation.

However, deeply binding pension funds, which are meant to address "emergency needs," with AI infrastructure that is still in the investment phase poses potential risks that cannot be ignored. While supporters argue that this is a necessary investment at the infrastructure level, analysts warn that exposing the public's retirement savings to the risks of a technology bubble without a clear profit roadmap could lead to painful lessons reminiscent of historical asset bubble bursts.

A Certain "Fit" Between Long-Term Capital and AI Ambitions

From the perspective of matching investment horizons, long-term funds and data centers indeed seem to be a "match made in heaven." The distant future envisioned by AI optimists coincides perfectly with the long-cycle characteristics of annuity businesses. The core logic of this view is: Although there may be misjudgments in the short term, an investment cycle spanning decades is sufficient to smooth out losses and ultimately achieve reliable returns.

Earlier this month, Masayoshi Son elaborated on his grand business vision. He calculated that if physical artificial intelligence can create 10% of global GDP in ten years—he believes that global GDP will reach about $200 trillion by then—then "even if you invest $10 trillion in the next ten years, you can recoup your investment in six months." Based on this logic, he posed the rhetorical question:

"Where is the bubble?"

There is no doubt that machine learning and artificial intelligence will profoundly impact the global economy in the coming decades. To achieve this goal, tech companies are currently undergoing large-scale financing to build the expensive data centers and power plants required. AI advocates believe that these projects will provide necessary infrastructure, including water, roads, bridges, and telecommunications networks. Therefore, using project financing to support AI infrastructure is seen as building tangible assets that underpin future investments.

Echoes of History: From the Japanese Bubble to Private Equity Concerns

Despite the grand vision, Bloomberg columnist Mayumi Negishi points out that when it comes to ordinary people's savings, the most conservative investment strategies should be adopted. She recalls the situation after the burst of the Japanese asset price bubble in the late 1990s: In less than four years, seven life insurance companies went bankrupt, severely impacting the "nest egg" of millions of policyholders. **

This is precisely why market observers are wary when SoftBank plans to introduce pension funds and the Canada Pension Plan Investment Board teams up with Goodman Group to make significant investments in data centers. Over the past decade, there have been numerous warning cases of Wall Street capital taking over life insurance companies.

A typical case is Jimmy Nappo, who diligently paid premiums for 17 years to secure the future for his wife and daughter. However, after his death, the insurance company he insured, owned by a private equity operator, went bankrupt due to a capital shortfall, leaving his widow still fighting for full compensation. These historical experiences indicate that when long-term savings are introduced into high-risk or highly leveraged investment tools, it is often the ordinary savers who ultimately bear the risk.

Huge Investments and Unverified Business Loops

Currently, the rationale for large-scale data center construction is primarily based on efficiency—larger scale leads to greater processing power and lower unit computing costs. However, the market has yet to see a fully validated roadmap that clearly outlines how to transition from the current investment frenzy to substantial profits that can cover the enormous costs.

Masayoshi Son's vision, while enticing, is criticized for not being based on real-world mathematical calculations. The industry is currently built on a massive assumption: the computing power of AI data centers will have value sufficient to cover capital costs and generate surplus, just like electricity. While life insurance companies and pension funds may control risks by investing in debt or preferred stocks to ensure priority compensation in adverse situations, the scale of investment is concerning.

Sam Altman of OpenAI has mentioned spending plans worth trillions of dollars and related financial instruments; Jensen Huang of NVIDIA has also emphasized the certainty of the AI era in a compelling manner. These statements are influencing investment managers in the financial sector. This inevitably brings to mind the excessive investment frenzy in infrastructure just before the burst of the internet bubble. While, in the long run, the technological bubble and the infrastructure it leaves behind may benefit society and improve human life, this does not mean that investing personal pensions in it is a wise move. For investors who primarily rely on pensions for their livelihood, avoiding such a grand gamble filled with uncertainty may be the more prudent choice