Forget the spy crap (although it’s bad enough, especially for Huang personally - although I don’t think the US DOJ would currently go after him, the potential Treason and violation of trade restrictions do not disappear so soon, and different people at the helm might yet try to uphold the law) - a lot more important is the fact that NVidia practically gave money to their clients to make them buy their overflow of silicon, just to make sure prices don’t go down. This means all numbers regarding their own revenue are untrustworthy and not representative of reality. Heck, it practically burns money to keep the money coming in and to get rid of overproduction ordered from TSMC.
Has anyone tried to remove all this virtual revenue from their balance sheets and then looked at how this reduced demand would affect prices and therefore the margin?
IMO it might also cause banks to say no to financing any more datacenter deals. A substantial part of NVidias revenue is practically coming out of their own pocket because of a missing costumer base that is willing to pay the prices asked. The income is a gamble, while the obligations are guaranteed. The GPUs bought from NVidia also act as a collateral for loans, meaning the perceived value of these cards are the reason they actually got their loans. If they are valued much too high because of NVidias little games… then the banks now sit on loans that are not secured by anything of value, much like 2008.
There’s also the issue with “Zombie Unicorns”, startups that are valued above one billion US$, but can’t go public to cash out because their valuation would not survive contact with the market, and the AI market is full of these - any investment made in the hopes of cashing out at the IPO might just not realize any gains because the potential payout is less than the investment + interest rate (Here the high inflation also comes into play). It is not clear yet which of those or how many might be undead (I think it is a sizable amount - there’s not enough income for all of them available, especially since they all sell somewhat the same with some modifications). Any respectable bank should try to minimize the impact of these facts on themselves.
Forget the spy crap (although it’s bad enough, especially for Huang personally - although I don’t think the US DOJ would currently go after him, the potential Treason and violation of trade restrictions do not disappear so soon, and different people at the helm might yet try to uphold the law) - a lot more important is the fact that NVidia practically gave money to their clients to make them buy their overflow of silicon, just to make sure prices don’t go down. This means all numbers regarding their own revenue are untrustworthy and not representative of reality. Heck, it practically burns money to keep the money coming in and to get rid of overproduction ordered from TSMC.
Has anyone tried to remove all this virtual revenue from their balance sheets and then looked at how this reduced demand would affect prices and therefore the margin?
Edit: https://finance.yahoo.com/markets/stocks/articles/michael-burry-real-nvidia-story-034429855.html talks about this: if financing slows, their commitments towards companies like CoreWeave (they are backstopping them, meaning they will practically pay them to do nothing if they can’t find costumers) do not disappear. This discrepancy will directly impact margin.
IMO it might also cause banks to say no to financing any more datacenter deals. A substantial part of NVidias revenue is practically coming out of their own pocket because of a missing costumer base that is willing to pay the prices asked. The income is a gamble, while the obligations are guaranteed. The GPUs bought from NVidia also act as a collateral for loans, meaning the perceived value of these cards are the reason they actually got their loans. If they are valued much too high because of NVidias little games… then the banks now sit on loans that are not secured by anything of value, much like 2008.
There’s also the issue with “Zombie Unicorns”, startups that are valued above one billion US$, but can’t go public to cash out because their valuation would not survive contact with the market, and the AI market is full of these - any investment made in the hopes of cashing out at the IPO might just not realize any gains because the potential payout is less than the investment + interest rate (Here the high inflation also comes into play). It is not clear yet which of those or how many might be undead (I think it is a sizable amount - there’s not enough income for all of them available, especially since they all sell somewhat the same with some modifications). Any respectable bank should try to minimize the impact of these facts on themselves.