AIG Didn't Lose a Bet. It Ran Out of Cash on a Bet It Was Still Winning.

In August 2007, the man running AIG's derivatives desk told investors that he could not imagine a scenario, "within any kind of realm of reason," that would cost the firm a single dollar on its credit default swaps.⁶ Thirteen months later the Federal Reserve made $85 billion available to keep AIG alive overnight.⁴ The easy lesson is that Joseph Cassano was a fool who lost the bet. But here is the part that should bother you: he may have been right about the dollar. Years later, even the Chicago Fed's own analysts could not cleanly say the bet itself was a loser. AIG did not fail because the bet went bad. It failed because it ran out of cash while the bet was still being argued over.

The official story is that AIG made a reckless $500 billion bet on subprime mortgages and detonated. Almost every load-bearing word of that is doing too much work. AIGFP did write swaps on over $500 billion of assets — but only about $78 billion of it was on the multi-sector CDOs that touched subprime at all.² A much larger slab was a quiet, boring trade nobody talks about.

The $500 billion that wasn't a housing bet

Of the $377 billion in net notional sitting in AIGFP's super senior portfolio at the end of September 2008, roughly $250 billion had nothing to do with American homeowners. AIG had written it for European banks that needed the swaps to lower their regulatory capital requirements — a paperwork trade, not a wager on housing.¹ Strip that out, and the famous half-trillion-dollar "bet" shrinks to the $78 billion that actually referenced subprime CDOs.² The popular accounts mash three different portfolios, measured on three different dates, into one monolithic number — and then act surprised that a number that large could blow up. It is the financial equivalent of describing a city by its area code.

So if the subprime slice was "only" $78 billion and the credit losses were genuinely disputed, what nearly killed a company with a balance sheet the size of a small nation? Not the bet. The collateral on the bet.

The killer wasn't loss. It was the margin call.

Here is the mechanism almost everyone skips, and it is the whole story. When you write a credit default swap, you typically agree to post collateral as the referenced assets lose market value — not when they actually default, but when their price falls. That distinction is everything. As mortgage-bond prices cratered in 2007 and 2008, AIG's counterparties came knocking for cash, regardless of whether a single underlying loan had yet failed. By late 2008 the multi-sector book showed unrealized losses of more than $33.9 billion, and AIG had already posted $22.4 billion of collateral against it.² You can see the panic accelerate in AIG's own numbers: the estimated unrealized loss on the super senior portfolio was put at $1.4 to $1.5 billion in November 2007, revised to roughly $6 billion by mid-February 2008, then to $11.5 billion two weeks later — the same portfolio, repriced three times in a hundred days.⁷ AIG didn't have to be wrong about the eventual losses. It just had to be unable to find tens of billions in cash on demand. It couldn't, and the auditors had already flagged a material weakness in its controls while the numbers were still moving.⁷

The other failure nobody mentions

And the swaps weren't even the only thing draining cash. The popular narrative treats AIGFP's derivatives as the sole cause, but the Congressional Research Service is blunt: AIG's exceptional losses "arose primarily from two sources" — the derivatives and a securities lending program run elsewhere in the firm.⁵ AIG had been lending out securities and reinvesting the cash collateral into, among other things, mortgage-backed paper. When borrowers wanted their cash back, AIG couldn't liquidate fast enough. That pool had a $58 billion U.S. balance outstanding in September 2008.⁵ It was a second, parallel run on the same firm — which is exactly why the Fed's first rescue, an $85 billion revolving credit facility approved by a unanimous Board on September 16, 2008, needed a separate $37.8 billion securities-lending facility bolted on three weeks later.⁴ One bailout had to stop two bleeds.

Notice the first entry on that timeline. AIGFP had already stopped writing new subprime swaps in early 2006 — and internal emails from 2005 show staff flagging "a huge amount of sub-prime mortgage exposure" as a worry.⁸ The desk that supposedly couldn't see risk had seen it, named it, and stopped adding to it years before the collapse. What it could not do was unwind what it already held without posting cash it didn't have.

But surely it was still a disaster for taxpayers?

The fair objection is that none of this absolves AIG. The risk controls were real failures — outside auditors found numerous weaknesses in risk management and internal controls⁸ — and a firm that has to be saved overnight by the central bank has, by any honest reckoning, failed. True. But the most repeated charge, that the rescue was a $182 billion taxpayer loss, is simply false. The Fed and Treasury made over $182 billion available, and the combined return on that commitment came to a positive $22.7 billion.³ The government wasn't just made whole; it turned a profit. That outcome is hard to reconcile with the story of a hopeless casino — and it fits perfectly with the story of a liquidity crisis at a firm whose assets were ultimately worth more than the panic price. You don't profit by bailing out something that was actually worthless. You profit by lending against something the market had momentarily refused to price.

Cassano told the crisis commission in 2010 that there would have been "few, if any, realized losses" on the contracts had they not been unwound in the bailout.⁶ He is an interested witness, and the claim is convenient. But it points at the genuinely unresolved heart of this story: AIG's near-failure was never cleanly a verdict on whether the bet was wrong. It was a verdict on what happens when you owe cash today against a price that has crashed, on assets whose true worth won't be known for years. The casino framing is satisfying because it lets us treat 2008 as a morality play about greed. The collateral framing is colder and more useful: a solvent firm can be killed by a margin call before it ever gets to find out if it was right. AIG didn't lose the hand. It got carried out of the room while it was still holding the cards.