Credit Head at Canada's Largest Pension Fund Warns of Bubble Signs in AI Financing
nashnova research
CPPIB, managing $617.6 billion in assets, warns that AI financing shows bubble-like overheating — circular funding structures and off-balance-sheet vehicles are amplifying hidden risk, and institutional investors need to reassess their AI exposure.
Why is one of the world's largest pension funds hitting the brakes on AI credit?
CPPIB credit head David Colla says AI-linked investments sit at just 3.5% of the credit portfolio, but the team is growing more cautious by the day.
His words were blunt: "You can't wake up one day and find AI is 25% of your book." This means → even for a long-term bull, position sizing matters more than directional conviction.
This reflects a broader signal: when the largest pools of patient capital start actively capping AI exposure, the market may be closer to consensus overheating than it thinks.
What structural risks are hiding inside AI financing?
Colla flags two core dangers. First, circular funding structures: Nvidia, Amazon, and Microsoft bankroll AI startups that depend on their own products. In plain terms = the lenders and the customers are the same group — if startups fail, the tech giants' own exposure balloons.
Second, off-balance-sheet financing: some AI firms set up shell entities to carry data-center construction debt that never appears on the parent's balance sheet. This means → lenders who only read the main financials may badly underestimate true leverage. Private AI developers are the riskiest blind spot.
Stack both structures together and a single startup failure could trigger a chain reaction well beyond one company.
If not AI credit, where is the money going?
During a roughly three-week market sell-off, CPPIB bought over $3 billion in discounted loans and bonds — a textbook contrarian move.
Colla named fund finance — lending to private-equity and credit funds — as the current priority. Moody's estimates the market has crossed $1 trillion, driven by the expansion of private credit and slower M&A forcing buyout funds to hold investments longer than planned.
CPPIB recently closed two large mezzanine deals in Europe — loans sitting between senior debt and equity — backed by loan portfolios, yielding roughly 4–5% above the benchmark rate.
Does fund finance itself carry risks?
Moody's has already warned: increasingly complex structures and layered borrowing in fund finance could amplify losses if underlying investments deteriorate.
CPPIB is also expanding into NAV lending — loans to private-equity funds collateralized by the estimated value of their holdings. In plain terms = fund managers who can't find a good exit borrow against the "paper value" of what they still own.
This reflects a tension: CPPIB flags structural risk in AI credit, yet the fund-finance space it favors faces a similar question — many of these products have never been tested through a full credit cycle.
What does CPPIB's own AI book look like?
It hasn't exited entirely. Current AI-linked positions include $265 million in CoreWeave, one of the largest AI-chip leasing firms, and $1.8 billion in Carlyle Aviation Partners.
For the fiscal year ending March 31, 2025, the credit portfolio returned 3.8%, with roughly 67% of assets deployed in the U.S.
Colla's summary: "There will be winners and losers along the way, and a lot of volatility." This means → the stance is not "avoid AI" but invest selectively, control total exposure, and watch the structures closely.
What does this mean for the market?
With OpenAI's fundraising shrinking and S&P downgrading Meta's credit outlook, CPPIB's caution is not an isolated call — large institutional capital is collectively repricing AI debt risk.
This means → financing costs for AI startups are likely heading higher, especially for firms reliant on circular funding or off-balance-sheet structures — they will face tighter due diligence and steeper rates.
In plain terms = the money is still there, but it is no longer cheap, and the bar to get it is rising.
市场有风险，内容仅供研究参考，不构成投资建议。