How AI Debt Hidden in Insurance Could Trigger the Next Crisis

The Trillion-Dollar Bet Hidden Inside Your Life Insurance

There is more than $1 trillion in artificial intelligence debt sitting somewhere in the global financial system. A significant portion of it is believed to be inside life insurance companies — the same companies that hold your annuities, your pension assets, and your retirement savings. And because of a regulatory clarification issued recently by the SEC, the disclosure rules designed after the 2008 financial crisis do not apply to most of this debt.

This is not a fringe theory. The structural mechanics are real, the dollar figures are documented, and the regulatory gap is confirmed. What remains unknown — deliberately so — is exactly how much AI-related risk is concentrated inside institutions that millions of Americans depend on for retirement income.

Here is a precise, data-grounded breakdown of how this structure works, why it echoes 2008, and what investors and policyholders should understand about where their money is actually going.

---

The AI Debt Machine: $1 Trillion and Growing

Building the infrastructure to run modern AI is extraordinarily expensive. Data centers, GPU clusters, and the power contracts required to run them cost tens of billions of dollars per project. The largest technology companies — Meta, Microsoft, Google, Amazon — are collectively spending more on AI infrastructure than their combined operating cash flows in some quarters. That gap is being filled with debt.

Google has raised over $30 billion in debt and issued a 100-year bond, meaning it won't mature until 2126. Microsoft and others have used similar instruments. According to Goldman Sachs estimates cited in research by Daniel Oliver of Myrmikan Capital, these companies carry approximately $1.5 trillion in lease commitments for AI infrastructure — but roughly $1 trillion of that does not appear on their published balance sheets.

Here is how that disappearing act works: instead of building a data center itself, a company like Microsoft signs a long-term lease contract with a third-party developer. That developer — not Microsoft — takes on the construction debt. The signed lease contract becomes the collateral. Because Microsoft is not the borrower, the liability does not show up as debt on Microsoft's financial statements. Investors scanning balance sheets for leverage risk will miss it entirely.

The result is a shadow debt market tied to AI that is larger than the entire investment-grade bond exposure to the banking sector. AI-linked debt now represents roughly 15% of the entire investment-grade bond market — a concentration that few outside institutional finance are tracking.

---

Private Equity Bought Your Insurance Company — And Then Lent You to Yourself

To understand how this AI debt connects to your insurance policy, you need to understand a structural shift that has been underway since 2009.

After the Federal Reserve cut interest rates to near zero following the financial crisis, life insurance companies faced a severe income problem. They had sold policies and annuities promising fixed returns — typically in the 3% to 5% range. When safe assets like US Treasuries and high-grade corporate bonds started yielding almost nothing, insurers could no longer generate enough income to honour those promises without taking on more risk. This is called "reaching for yield," and it opened the door for private equity.

Private equity firms recognised that insurance companies hold something extraordinarily valuable: permanent capital. Unlike a typical PE fund with a 10-year lifespan, an insurance company collects premiums every month for decades. That float — the pool of invested premiums — is stable, long-duration, and legally required to be invested in something. From 2009 to 2024, private equity ownership of life insurance companies grew from near zero to more than $700 billion in assets across 134 insurers. Total estimates for insurance assets under private equity control now exceed $1.5 trillion.

The conflict of interest built into this arrangement is structural and significant:

• Apollo Global Management owns the life insurer Athene. Apollo originates private credit deals. Athene buys those deals. Approximately $227 billion of Apollo-originated assets sit inside Athene's balance sheet.
• KKR owns Global Atlantic.
• Blackstone holds stakes in multiple insurers.
• Brookfield owns American National.

In each case, the same firm that creates the loans also controls the insurance company that purchases them. The firm earns origination fees for making the loan and management fees for running the insurer. Policyholders earn a fixed return regardless of how the underlying assets perform. If the assets deteriorate, the policyholder's contractual return is not immediately affected — until losses become large enough to threaten solvency itself.

The analogy that clarifies this most sharply: imagine your bank made a loan to a struggling business, then stuffed that loan directly into your 401(k) — collecting fees at both ends, while you absorbed the credit risk without knowing it.

---

Why the 2008 Rulebook Doesn't Apply to AI Securitisation

After the 2008 financial crisis, regulators introduced two critical rules for securitised products:

1. Regulation AB — requires detailed asset-level disclosure for asset-backed securities. If you sell a bundle of loans to investors, you must disclose what's inside.
2. Regulation RR (Dodd-Frank risk retention) — requires the originator to retain at least 5% of any securitised deal. This "skin in the game" provision is meant to discourage the originate-to-distribute model that flooded the pre-2008 market with poorly underwritten loans.

Both rules target asset-backed securities. A data center, however, is not a financial asset in the traditional sense — it is a physical building. When a law firm working on AI infrastructure securitisations asked the SEC to clarify whether these rules applied to data centre debt deals, the SEC confirmed they do not. These structures are not classified as asset-backed securities.

The consequences are direct:

• No mandatory disclosure of what is inside these bundles of AI debt.
• No risk retention requirement, meaning originators can structure and sell deals with zero skin in the game.
• No standardised reporting on loan performance or underlying collateral quality.

This is precisely the opacity that made 2008 so destructive. Nobody knew what was inside mortgage-backed securities until the losses started cascading. With AI infrastructure debt, we are in a similar position — except the holding institutions are not banks. They are life insurers, and the backstop for life insurers is not the FDIC. It is the State Guarantee Association system.

---

The Insurance Backstop: How a Private Risk Becomes a Public Liability

Most policyholders are unaware of how the insurance safety net actually works. The State Guarantee Association (SGA) is a system where, if a licensed life insurance company fails, all remaining licensed insurers in that state are assessed — forced to contribute — to cover the failed company's obligations to policyholders, up to state-specific limits (typically $250,000 to $500,000 per policyholder).

Here is the critical detail: those assessed companies typically receive the money back as a credit against their state premium taxes. Which means the state — funded by taxpayers — ultimately absorbs the cost. This is a de facto public backstop for private insurance solvency, without the explicit congressional approval a bank bailout would require.

The implication, if the AI debt concentrated inside insurance companies deteriorates significantly:

• Policyholders face first exposure as insurer solvency weakens.
• Competing insurers are assessed under the SGA system.
• State governments reimburse those assessments via tax credits.
• Taxpayers fund those state governments.

The losses originate in private equity-engineered AI debt structures. They terminate on public balance sheets. The firms that designed and profited from the structure are insulated at every step by fees already collected.

A further complication: a significant portion of this insurance-linked AI debt has reportedly been moved offshore — particularly to Bermuda, where disclosure requirements are minimal and regulatory oversight is lighter than in US jurisdictions. This makes accurate accounting of total system exposure effectively impossible from the outside.

---

What This Means for Your Retirement Accounts and Investments

If you hold any of the following, this structure is directly relevant to your financial situation:

• Fixed annuities or indexed annuities from any major insurer
• Life insurance policies with a cash value component
• Pension plans with assets managed by private equity-linked managers
• Bond funds with investment-grade corporate or private credit exposure

None of this means a collapse is imminent or inevitable. AI infrastructure may generate the long-term cash flows needed to service this debt. The technology buildout, while expensive, could prove as transformative and value-generative as the internet infrastructure boom. That case exists and has credible advocates.

But the risk profile of the system is not being accurately communicated to the people who bear it. Policyholders signing annuity contracts are not being told that their premiums may be funding 24-year bonds tied to data centers, originated by the same private equity firm that owns their insurer.

Investors and policyholders who want to act prudently on this information might consider:

• Reviewing the financial strength ratings of their insurance company (AM Best, S&P, Moody's) and monitoring them annually.
• Understanding what assets back their annuity or policy — many insurers publish statutory filings that give partial insight.
• Diversifying retirement income sources rather than concentrating in products from a single large insurer.
• Staying within SGA coverage limits per insurer per state, particularly for annuities.
• Following SEC rulemaking on private credit and insurance company investment disclosures — this is an active regulatory area.

---

The 2008 Parallel: What's Different, What's the Same

The structural similarities to 2008 are not superficial. Both episodes share: opaque bundling of debt, regulatory gaps that exempt key instruments from disclosure rules, concentrated risk in institutions with public backstops, and originate-to-distribute incentives that sever the link between loan quality and lender accountability.

The differences are also real. The underlying asset — AI infrastructure — has genuine long-term demand, unlike the overbuilt housing stock of the mid-2000s. The counterparties signing the lease contracts (Microsoft, Google, Meta) are investment-grade credits with strong balance sheets, not subprime mortgage borrowers. And the insurance regulators, though operating under lighter rules than banking regulators, do conduct solvency examinations.

But the scale is comparable. AI-linked debt at roughly 15% of the investment-grade bond market is a systemic concentration. The insurance industry holds approximately $849 billion in private credit — around 42% of that entire market. These are not marginal numbers.

History suggests that systemic risks embedded in opaque structures are not typically resolved by the market self-correcting gracefully. They are resolved when something breaks and the true ownership of losses becomes unavoidable. At that point, the question of who holds the bag is already answered.

---

Conclusion: Ask the Questions Before the Losses Force the Answers

The core issue here is not that AI is overvalued, or that private equity is malicious, or that insurance companies are insolvent. The core issue is that a chain of financial engineering has moved risk — systematically and deliberately — from the entities that profit from it to the entities that cannot easily track or exit it. Those entities are pension funds, annuity holders, and ultimately state taxpayers.

That shift has happened before. It ended badly. The fact that today's version is more sophisticated and involves physical buildings rather than mortgage pools does not make the underlying logic less familiar.

The three questions worth asking right now:

1. Who issued your annuity or life insurance policy, and who owns that company? If the answer is a large private equity firm, investigate what assets back your policy.
2. What is the credit quality and transparency of your fixed income or private credit allocations? The disclosure gap in AI infrastructure debt is real and documented.
3. Is your retirement income concentrated in a single insurer, or diversified across multiple institutions and asset types? Concentration is the variable most within an individual's control.

The trillion dollars of AI risk in the financial system may never cause a crisis. But the people bearing that risk deserve to know they are bearing it.

---

This article draws on research by Daniel Oliver of Myrmikan Capital and publicly available data on insurance industry asset allocation, private equity ownership structures, and SEC regulatory guidance on AI infrastructure securitisation.

This article is for informational purposes only and does not constitute financial advice. Always consult a qualified financial professional before making investment decisions.

---

Frequently Asked Questions

What is the State Guarantee Association and does it fully protect my annuity? The State Guarantee Association (SGA) is a state-level backstop that requires solvent insurance companies to cover policyholders of a failed insurer. Coverage limits vary by state but typically range from $250,000 to $500,000 per policyholder for annuity contracts. Critically, this is not a federal guarantee like FDIC deposit insurance, and the costs are ultimately absorbed through state tax credits — meaning taxpayers fund the backstop indirectly. If you hold annuities above these thresholds, the excess is not protected.

Why don't the post-2008 securitisation rules apply to AI infrastructure debt? The SEC recently confirmed that AI data centre financing structures do not qualify as asset-backed securities under existing regulations, because a data centre is a physical asset rather than a self-amortising financial asset like a mortgage or auto loan. As a result, Regulation AB disclosure requirements and Dodd-Frank risk retention rules (which require originators to hold 5% of securitised deals) do not apply. This creates a disclosure gap similar to the opacity that characterised pre-2008 mortgage securitisation markets.

How does a private equity firm owning an insurance company create a conflict of interest? When a private equity firm both originates loans (through its credit arm) and owns the insurance company that buys those loans (through its insurer subsidiary), it sits on both sides of the transaction. The PE firm collects origination fees and management fees; the insurance policyholders bear the credit risk. If the loans perform badly, the PE firm has already been paid — it is the policyholders' fixed returns and ultimately their capital that is at risk. This structure gives the originator limited incentive to prioritise loan quality over loan volume.

Should I move my money out of annuities or life insurance because of this? That is a decision requiring personalised financial advice based on your full situation. What this analysis does suggest is that policyholders should: check the AM Best or S&P financial strength rating of their insurer annually; understand whether that insurer is owned by a private equity firm with affiliated credit operations; ensure their annuity balances stay within SGA coverage limits per insurer; and consider diversifying retirement income across multiple institutions rather than concentrating in a single product or provider. Do not make changes based on any single article — consult a qualified financial adviser.