> To me, that seems to be a requirement for the computing industry for a long time.

Sure, but they have a market cap of 5 trillion. It's about 10x that of AMD, which also sells similar silicon (and isn't in any distress). It's more than Apple, Google, and Microsoft - and these companies historically found ways to make more money than the vendors they buy chips from.

The problem isn't that Nvidia doesn't have good fundamentals or good products, it's that the market is expecting miracles.

In the case of Nvidia, the funny thing is that their high valuations started not with AI, but with cryptocurrencies. Just never really came down - they coasted from a silly hype cycle to a more substantive one. Ten years ago, NVDA wasn't an interesting stock at all.

Stock values go up when people buying stocks expect them to go up.

Stock values go down when people holding stocks expect them to go down.

Why would you hold a stock if you think it should go down? If you think the stock is valuable but in the near term should go down, why not sell and then buy in increments as it goes down?

Because stocks have a tendency to go up even when they should be going down. And when you decide that it probably isn't going down, it will go down. Timing the market isn't a reliable way of wealth generation. Long term investing is.

Because not everyone is in it for the short term.

Tax reasons might be one as well, long term capital gains are taxed less.

There are few investors that can spend the time it takes to be active like that.

Most people buying individual stocks are better of buying ETFs anyway.

In the end it's a choice on what to spend your time on.

Indeed. Since most stock are mostly held by institutional investors, prices are heavily guided by sentiment.

Except when they aren’t, see GameStop and Beyond Meat.

Prices are heavily guided by sentiment. Nobody said sentiment HAS to be tied to the entity's fundamentals. GameStop stock moved due to sentiment external to the entity itself.

Nvidia is already doing that under MGX. They also offer their Grace CPUs on that platform.

The problem is not demand going away. No margin in a late stage company goes unassailed for long. Intel has nothing to lose. AMD has everything to gain. Untold other players are finding oxygen in various places. Nvidia is smart to use their spotlight as long as it lasts, but in their pitch, they're saying, "this will put you ahead," not "this will last forever."

Google?

90% search is not 90% ad revenue. If you want to compete with Google, it's easier to make a different kind of popular platform and then sell ads than to compete head-to-head for traffic on search only to... sell ads.

> The problem is not demand going away.

The problem for Nvidia is when demand doesn't continue to increase as much as expected.

Unless they have managed to hide a lot of leverage and debt (not investment), that's a problem for their shareholders if they have priced in future growth.

To avoid looping this conversation for the next 2-3 years, we need to get to higher PEs to match the dot com bubble. It's just hard to have a long bubble that collapses without debt and leverage. You need something that compels the market to correct or else the long bubble just stays long. Suddenly I understand why shorting is essential in breaking the back of long bubbles and stabilizing market dynamics.

Yes, but their moat is not unassailable. If/when the market for massive parallel computations becomes truly competitive, the combined market cap of all companies in it will likely be smaller than what NVidia currently has. NV margins are insane, and are only possible because they have an effective monopoly.

> The next logical step would be to get into CPUs, to become a fully integrated computing solutions provider.

They already tried it 5 years ago [1][2] but it was promptly blocked by regulators.

[1]: https://nvidianews.nvidia.com/news/nvidia-to-acquire-arm-for...

[2]: HN discussion https://news.ycombinator.com/item?id=24464807

They were prevented from buying out ARM.

They can always license their IP and make products out of that.

Some of their systems most likely have such chips already in them.

It wouldn't surprise me if they are looking into making their own RISC-V CPUs, since their coprocessor already is build on it already. They also are porting CUDA to RISC-V (specifically RVA23 is their minimum supported) so it would somewhat make sense to do it.

This will sound like I'm joking, but I'm not. It seems like with this administration, having the regulators reverse their decision wouldn't be that hard, especially with a "donation" to the ballroom or something along those lines.

I don't think the UK is interested in a ballroom. The UK blocked that deal, not the US.