Their major risks (excluding future competition from China and elsewhere) is, after the crash, receivable issues with bankrupted companies and competing against "discovered" overstock of their chips being liquidated by others.

I think that by far their biggest risk is simply competition. Their clients are gigantic: several have the means, opportunity, and motive to invent their own substitutes and get TMSC to fabricate them so that they can capture the billions in profit that Nvidia is making on their purchases. The software ecosystem is a nice barrier to entry, but far from insurmountable for the tech giants. So I don't think they have a barrier to competitive entry, just a very high cost of competitive entry, in both money and elapsed time. Anyone meeting that cost could drive their returns (measured as a percent of that cost) down to the typical return on capital.

The only unusual situation here is that such a new competitive entrant might be captive, and pull away only the business from their owner. In other words, if (say) Microsoft or whoever designs their own chips for their next generation data centres, they might not sell those chips to the other clients of Nvidia. That outcome would considerably reduce the risk to Nvidia's future profits.

Jim