00:00Joining us now is Goldman Sachs Semis analyst Jim Snyder. He has a buy rating on the stock with a
00:04$250 price target.
00:07So, Jim, if all expectations, yours including today, is a beat and a raise and we've already baked in 80
00:13% sales growth,
00:14what does it take to get the stock to have a positive reaction, to actually pop after they report results?
00:20Yeah, it's a great question on top of many investors' minds. I think we've had down reaction that last three
00:24out of four quarters for the stock.
00:27Partly not because the numbers were not good, because the numbers have been very, very good, over 80% growth,
00:31as you said.
00:33But I think there are some questions that have, some investors have questioned the sort of sustainability and durability of
00:38those earnings.
00:39I actually think that we're going to not only have a beat and raise for this quarter, but that's in
00:44fact sustainable into next year.
00:46Our estimates for next year are over 30% above the street.
00:49And I think there's been a concern among the hyperscalers and model providers about the sustainability and durability of their
00:56spending.
00:56And that kind of becomes a concern about NVIDIA as well.
01:00I think the way I'm seeing it is, and we wrote out another note a couple weeks ago on this
01:06topic, we actually think that as the costs come down for the tokens
01:11that everyone is consuming very quickly by about 70% a year or better, that's being given by the chip
01:19providers like NVIDIA and Broadcom.
01:20Sorry, how are the costs of the tokens coming down? Is this just efficiencies that these companies are eking out?
01:25Yeah, it's architectural efficiencies, both in terms of the fundamental chip compute, but also the networking between all of the
01:31different chips in the data center or cluster within a data center.
01:33So, I mean, one concern that Danny's had lately is inference costs.
01:38Yes, well, Uber ran through their entire year's budget in just a couple months of inference costs.
01:42But that kind of concern will dissipate, I guess, over the coming years.
01:47We think so. And, in fact, we show in our recent report, these token costs are coming down by the
01:54chip providers by about 70% per year.
01:56That's very rapid. In fact, that's much more rapid than we saw during Moore's Law, which was 10, 15%
02:03kind of annualized cost declines.
02:04This is much quicker than that.
02:06And so I think that really what it's really doing is it's driving what has been sort of rather sort
02:13of a tepid gross margin for the hyperscalers and model providers where now token costs are actually coming down faster
02:21than token price.
02:22And those those two things cross token price cost falls below token price.
02:28You start to get positive gross margin and better unit economics for the hyperscalers.
02:33That makes their spending more sustainable.
02:34I have been thinking a lot about NVIDIA chip sales to China over the past year.
02:42And I think it's pretty clear that the Chinese want to avoid taking NVIDIA chips, giving them more revenue and
02:51obviously foster growth in their own chip companies at home.
02:54So that takes the, you know, the 50 billion dollar TAM story off the table, at least to some extent.
03:01How can their total addressable market grow from here?
03:04It's more around the core U.S.-based hyperscalers and model providers.
03:09In fact, today spending among that cohort of companies is more than 20 times today's opportunity in China.
03:16Not that we don't believe that China opportunity long term will be very significant, but today it's quite small in
03:21comparison.
03:21So if you look at where NVIDIA is trading, I mean, it's still, you know, 26.7 times forward earnings.
03:28It's not cheap, but it's it is cheaper relative to where they have been in history as recently as the
03:32end of last year.
03:33It was trading at 31 back in 2023.
03:35It was trading at 43 times earnings.
03:38What's changed for NVIDIA that it's not trading as richly priced earnings as it has in the past?
03:44Well, I think part of it is law of large numbers.
03:46Part of it is law of large stock.
03:48But also there's been some concern, as I said before, among around the sustainability of that future spending and whether
03:55you can really grow at these rates on top of these very, very large numbers we're at today.
03:59So, you know, we're have close to 80 percent revenue growth again in 2027 off of these numbers.
04:05And as I said, we're 30 percent above the street.
04:07So if that turns out to be true, the stock is actually much more attractively priced than that.
Comments