AI Pulse

Is NVIDIA becoming the AIG of the AI boom?

AI Markets 374 min read

A sharp analogy landed in an FT column: the chip giant's role in the AI system is that of a quasi-central bank, echoing AIG before the subprime bubble burst. Whether it holds depends on what exactly is being compared.

Where the comparison comes from

The FT column argues that NVIDIA occupies a position in the AI economy close to a quasi-central bank, reminiscent of AIG's position before the subprime mortgage bubble.

The point is not size but position. When the liquidity, credit and expectations of an entire system route through a single node, that node starts to behave like a quasi-public institution.

What AIG got wrong

AIG sold credit default swaps that provided implicit guarantees on mortgage securities, collecting steady premiums while concentrating the payout obligations of extreme scenarios on its own balance sheet. The business was not the error; the mismatch between collateral and capital buffers and the pace of exposure growth was.

When the insured assets deteriorated together, AIG stopped being the party collecting premiums and became the party that had to pay immediately.

Where NVIDIA looks similar

  • It sits at the centre of the AI capital loop: compute supply, customer expansion, ecosystem investment
  • If downstream demand weakens, the shock does not disperse evenly — it flows back to that centre
  • The phrase quasi-central bank itself implies a kind of implicit backstop

But the differences are fundamental

NVIDIA sells physical chips, and revenue corresponds to real deliveries and cash flow. AIG's exposure was derivative notional, and the loss profile in an extreme scenario is not comparable. Treating the analogy as a forecast underrates the cash-generating capacity of a hardware business.

A more precise framing: NVIDIA carries cyclical risk; AIG carried tail risk. The first hurts, the second kills.

The real risk is structural

  • Customer concentration: a handful of very large buyers set the order rhythm, so any slowdown gets amplified
  • Circular investment: vendors invest in customers who then buy more, widening the gap between reported revenue and end demand
  • Debt-funded purchasing: when buyers borrow to buy compute, a shift in funding costs lands directly on orders

The last one is already visible: a $32 billion chip-linked debt deal was priced at a yield above 8%.

What to watch

  • Trends in receivables and inventory turnover
  • Disclosures on customer financing and circular arrangements
  • Issue yields and subscription levels on AI infrastructure debt
  • How credit markets price NVIDIA itself

Nothing here is investment advice.

Related coverage