- 3 hours ago
Category
🤖
TechTranscript
00:00NVIDIA's earnings failing to impress investors despite beating the already very high expectations. The results almost overshadowed by
00:07headlines related to this year's mega IPOs with SpaceX unveiling its filing. Let's discuss with Gil Luria head of technology
00:16research at D.A. Davidson and Joe Kaiser. Switchyard Partners CEO and founder. Joe has built and backed companies through
00:23multiple
00:25IPOs, worked in solar, worked in AI. And Gil, let me start with you since you're playing the away game
00:33here
00:34today. What do you make of NVIDIA earnings? It seems like they're crushing it, but the bar is just so
00:41high
00:41and they've already gotten so big. That's right. They did crush it. They'll continue to crush it.
00:48The highlight of the earnings yesterday was Jensen Wang dunking on the competition. He made a
00:54point to dunk on AMD and Broadcom by saying there's an entire segment of his business where they don't
01:00have any competition apart from the companies that dabble in their own chips. The entire rest of the
01:05market is 100 percent NVIDIA. Then he proceeded to dunk on the CPU companies, Intel and AMD, by talking
01:12about the fact that in a year they're going to grow their standalone CPU business from nothing to 20
01:19billion. And then he ended by dunking on Cerebris by saying that the market for hyperspeed inference is
01:26actually tiny in spite of the fact that he's also in that market. So NVIDIA is doing really well.
01:33It's the core of the AI trade and will continue to be the core of the AI trade.
01:38The Jalen Brunson of corporate America just dumping on everyone. Joe, if I can put you in, I actually want
01:43to
01:43kind of relate this to SpaceX because it's a point that Mandeep Singh made of Bloomberg Intelligence that I thought
01:47was good. You have SpaceX valued at two trillion dollars. It's going to be priced theoretically if it gets there
01:54at 80
01:54times sales. It's going at something like 20 percent. Whereas you have NVIDIA priced at 17 times sales or forward
02:01earnings and it's growing at 85 percent. How do you digest with these giant tech companies coming to market about
02:07what
02:07you're actually willing to pay. Danny, it's astounding. And then you put OpenAI and Anthropic on the heels of that
02:15and their
02:15growth rates are extraordinary as well. I mean, just highlighting something else that Gil or adding on to what Gil
02:23was talking
02:23about, like the networking business inside of NVIDIA was lost in all the conversation. They're at $15 billion in revenue
02:31for
02:31their networking equipment. That is puts them on a run rate equal to Qualcomm. I mean, you just and that's
02:37just an also ran
02:38business inside of NVIDIA. So Jensen talked about the the AI factory at the GDC and it's coming to life,
02:45by the way. So they're
02:47expected to have $370 billion in annual sales. So at 5.4 trillion. They're only trading at 14 and a
02:55half times. Thank you for that
02:57sales and about 20 times forward earnings. Right on a 12 month basis compared to SpaceX, which would be 80
03:03times sales. Joe, Jensen continues to
03:06dangle, though, this trillion dollar figure for revenue. Right. Is that 25 through 27. Are they on track to hit
03:12that? All signs point to
03:15yes, Matt. I mean, it is astounding that Jensen can just trot out this crazy number that year over year
03:21is 100 percent bigger than he
03:23said the year before. And he's on a runway to hit it. And Matt, I think the other interesting component
03:28here about their moat is the
03:29lockdown that they have on TMC production. The moat is not silicon. The moat is the packaging. And as a
03:36consequence, it's clean
03:37rooms is where the packaging occurs. They are consuming almost two thirds of all the package chips coming out of
03:45TSMC. So if you're
03:47trying to get Google chips, if you're trying to get training and chips from Amazon, how are they going to
03:53make them? They all get made in the
03:55same place, which is the most interesting thing there. Yeah. Yeah, exactly. And that and there's the supply chain risk
04:00within that, too. Gil, if I can bring you back into the conversation, I know you have some thoughts on
04:06the SpaceX IPO, too.
04:08And I think one of the ones I'm most interested in hearing is the twenty eight and a half trillion
04:13dollar
04:14supposed TAM total addressable market. What what did you make of that? How do you think we can get to
04:19that kind of figure for
04:20SpaceX and can we? Yeah, I mean, the nature of SpaceX is that it's really about the dream, right?
04:28The mission statement includes statements about humanity becoming multi-planetary, extending the light
04:36of consciousness into the stars. So that sounds like multi-trillion dollars to me. I actually had this
04:43conversation with Elon Musk 13 years ago. I asked him about taking SpaceX public. Tesla was public at
04:52the time. It was only a five billion market cap and it was already giving him headaches. So he said,
04:57no, I think
04:57I'm going to keep SpaceX private for a while. So here we are 13 years later and they're both going
05:04to be
05:04trillion dollar companies. So believe Elon when he says the opportunity is big. Joe, can they get there to a
05:11two trillion dollar valuation? I mean, Tesla is valued at one and a half trillion dollars and they don't
05:19even make and sell that many cars. Right. It's all about the dream for Tesla, too. And they seem to
05:25be able to
05:25support that. Matt, the dream is grounded in reality here, though, because Tesla's our Tesla. Sorry.
05:31SpaceX is already launching infrastructure into space right now to support this data center in space model. So it is
05:38actually
05:39not really enough. Right. Starship needs to be launching 100 to 150 tons of payload and they keep blowing up.
05:46So they're
05:46not launching any commercial payload thus far. And if they did get customers to put stuff into those, they could
05:52only take up like 35 to
05:5450 tons. Sure. The one thing that the last couple of decades have proven to us, Matt, is don't bet
06:00against Elon. So the
06:02the tactical technology solutions are will be solved. I think, Gil, for me, one of the head scratchers is
06:09their XAI business because they are burning through cash. And in fact, SpaceX has really big GPU clusters, but
06:17they're renting more of that to Anthropic than what they're actually using themselves. How should we think
06:23about the AI business behind SpaceX and its potential growth?
06:28Yeah. So up until now, this one tells us that it's been a huge
06:33money drain on Tesla, but that turns around because they're going to start charging Anthropic
06:38$15 billion a year for that capacity. What happened is Elon's great at building these Colossus data
06:45centers, but Grok wasn't popular, hasn't been popular enough in order to fill them with inference
06:51demand. So he has this excess demand that he can sell to Anthropic, pay for the investment while
06:58Grok gets better. And at which point he can give Anthropic a 90 day notice to kick him out and
07:05use
07:05that capacity for Grok. So bigger picture, understand what Elon does. He divides the business into three
07:14parts. There's the money making business today. For Tesla, that's cars. For SpaceX, that's Starlink.
07:21There's the business that's on the come. For Tesla, that's RoboTaxi. For SpaceX, it's data centers in
07:27space. And then there's the dream. For Tesla, that's Optimus Robots. For SpaceX, it's Life on Mars. All these
07:35things are related, but all these things allow him to get a valuation that's much bigger than the existing
07:42business would justify because there's a shorter term promise and a longer term promise that investors
07:49can get excited. But Gil, if we have a company, I mean, I don't know what the exact estimates are
07:55for
07:55the full year, but let's say $20 billion in revenue and $20 billion in losses. If a company that loses
08:06$20 billion on an annual basis can be a $2 trillion valuation, don't you start to sell this market
08:11because that's the top? Well, there's three really exciting companies and they're all going
08:17to IPO this year. And all three of them lose tremendous amount of money, but all three of them
08:22have a TAM that is in the multiple trillions. That's obviously these guys, OpenAI and Anthropic.
08:30And the market is hungry for this type of growth. The market believes these companies are going to
08:36be the engine of growth going forward. Now, what's going to happen is these IPOs are going to take
08:41the oxygen out of the roof. They're going to make everybody else look bad because of their growth
08:45rate and the size of their opportunity. So investors will be willing to accept these losses as long as
08:52the trajectory for growth and the size of the opportunity is great. That's part of the dream
08:57that investors have right now. And that's the opportunity they're looking for, for having
09:02a really big opportunity. These are the three companies that are going to provide that to the
09:07market. Joe, I got to ask you about OpenAI because I don't have a big picture analytical view,
09:16but the people that I talk to on the ground who are using AI deeply have switched to Anthropic,
09:24right? Joe Weisenthal, when he's building code for OddLots, he's doing it with Claude.
09:29Right.
09:30How is OpenAI doing in the sort of Gemini, Claude, Grok, ChatGPT horse race?
09:36Yeah, man. So obviously most of this data is private. So it's a matter of piecing it together.
09:42And OpenAI's growth rate from 23 to 25 is pretty extraordinary. It went from $2 billion to $20
09:48billion in revenue.
09:49And they were the only game in town, though.
09:50Extraordinary. But a couple months ago, we saw it. The flip occurred with the release of the new
09:56Anthropic tools, Code in particular. And now Code has double the users of Codex, which is
10:04extraordinary. So if you think about what software are engineers using to build product, that is a
10:10two-horse horse race. And Code is winning. Anthropic is winning the race today. The Anthropic's
10:16growth rate is much faster at this point than OpenAI's. And I'm sure that's a little bit of
10:22Sam's motivation to get out of the gate sooner.
10:25I was going to ask you that, Joe. As someone who has taken companies public, how important is it for
10:30OpenAI to get there before Anthropic or even SpaceX for fear that oxygen gets taken out?
10:38Sure, sure, sure. Well, I think OpenAI just needs the capital at full stop. I mean, they are talking about
10:44$600 billion of infrastructure deployment through the rest of the decade. So they just need the capital. That is
10:50definitely the primary motivation. But SpaceX may be less, but they must get out before Anthropic.
10:57Gil, let me put a hypothetical to you. If OpenAI continues to lose this race against Anthropic,
11:05is that a problem for the entire multi-trillion dollar circular economy that's all based on OpenAI winning?
11:15If Anthropic runs away with it, yes. But just over the last six months, we've gone from OpenAI having a
11:22lead that looked insurmountable to Google. Gemini having a lead that looked insurmountable to Anthropic
11:29taking the lead. When we're actually looking at the numbers coming out of usage in Enterprise,
11:36the usage of OpenAI at the Enterprise is actually almost as big as Anthropic. Anthropic grew it very
11:42quickly, but most companies actually want to use both tools. And part of the reason is that Anthropic
11:48didn't have the compute to deliver to this demand. That's why they're scrambling to get compute
11:53everywhere, where OpenAI was ahead of the curve in securing compute. So if they have a slightly better
12:00model, if they can deliver more compute, they'll come right back into the lead. So it's too early to
12:06call, not to mention Google Gemini over the last couple of days presented a very compelling case why
12:12enterprises should use that model and enterprises are keeping Google into the mix. The race is still
12:19completely wide open. And to Joe's point, that's why OpenAI needs to go get that capital,
12:24pay for the compute so they can deliver it so they can stay in the race.
Comments