The artificial intelligence boom is reshaping more than just the technology sector. Behind the scenes, the massive buildout of data centers that powers AI is being financed in a way that could change how everyday investors experience risk — and not necessarily for the better.
As pension funds and 401(k) plans pour money into so-called semi-liquid funds, a growing share of that money is ending up in private credit, a market where prices are often set by models rather than by real-time trading. That means the risk is being priced in a way that can lag reality, and everyday savers may be the last to know when prices catch up.
What's going on here?
Big Tech's AI data-center buildout is increasingly being funded with private credit, so more of the risk is being priced by models, not by real-time markets. Private credit refers to loans made by non-bank lenders, such as private equity firms or specialized credit funds, that are not traded on public exchanges. Instead of a live market price, these loans are valued using models that estimate what they would be worth.
This is not a niche corner of finance anymore. A Reuters column by Altaf Kassam of Reference Portfolio Partners notes that AI-linked private lending is now big enough to shape how investors experience risk, not just returns. The OECD estimates that AI-related deals made up 34% of global private credit transactions in 2025. The Bank of England expects about $700 billion of data-center spending over 2026-28 to be financed this way, in a market the Financial Stability Board sized at $1.5-2 trillion at the end of 2024.
To put that in perspective, the entire private credit market is roughly the size of the U.S. high-yield bond market. And a significant slice of it is now tied to the fortunes of AI infrastructure, which is itself a bet on future demand for computing power.
What does this mean for investors?
For most people, the connection is indirect but important. Pension funds and 401(k) plans increasingly allocate to private credit through semi-liquid funds, which allow investors to redeem their money periodically, but not daily like a mutual fund. These funds often use “smoothed” model-based marks, meaning they adjust valuations gradually rather than in real time.
The problem: if the underlying loans lose value quickly — say, because a data-center project runs into trouble or interest rates spike — the fund's stated value may not reflect that immediately. By the time the model catches up, some investors may have already redeemed at a higher price, leaving those who stay in the fund to absorb the losses. In other words, everyday savers could be last to learn when prices catch up.
This is not a prediction of a crash, but it is a reminder that private credit's appeal — higher yields and lower volatility — comes with a trade-off: less transparency and less liquidity. The smooth ride can turn bumpy when the models finally adjust.
Why data centers are the new frontier
Data centers are the physical backbone of AI. They house the servers that train and run large language models, and they consume enormous amounts of electricity and water. Tech giants like Microsoft, Amazon, and Google are spending tens of billions of dollars each year to build them out.
That spending has created a boom for companies that supply the equipment and power. For example, Siemens has raised its growth target thanks to AI data center demand, and Siemens Energy's order backlog has hit records. Even Microsoft and Chevron have teamed up to power data centers without subsidies.
But the financing side is less visible. Traditional bank loans are often too big or too risky for a single lender, so private credit funds step in. These funds can move quickly and structure deals flexibly, but they also take on more risk. The result is that a growing share of the AI boom is being financed by a market that is less regulated and less transparent than public markets.
What to watch next
Investors should keep an eye on a few things. First, how much of their pension or 401(k) is actually in private credit — it may be more than they think, given the rise of semi-liquid funds. Second, the pace of data-center spending: if it slows, the loans backing that spending could come under pressure. Third, any signs of stress in private credit, such as rising default rates or funds restricting redemptions.
For now, the AI data-center boom is still going strong, and private credit is a key engine. But the shift to model-based pricing means that the risk is being smoothed over, and that smoothing can hide problems until they are harder to fix. As always, diversification and understanding what you own are the best defenses.
