00:00Our next guest believes markets are still digesting the largest capex surge in history,
00:05driven by demand for AI compute. Todd Olson, Parnassus Investments, Chief Investment Officer
00:09and Portfolio Manager is with us here in San Francisco on set. I want to start with the
00:15signal, I suppose, that comes from that report Rachel just gave. Anthropic has a real revenue
00:20number and an updated ARR number and lots of different names are moving this morning. Chip
00:26names, those that are invested in it, other software names. Why do you think that is?
00:31Well, thanks for having me, Ed. And after 31 years of being in the business, we're just
00:35at a stage now where the physical world just can't keep up with the digital world. That's
00:39creating incredible bottlenecks and that's widely known. And I think now you're seeing this
00:44wave of liquidity. And then we saw NVIDIA last week with the 500 billion announcement, Anthropic,
00:50SpaceX, OpenAI. So we're basically hitting the stage where a wave of liquidity is going
00:56is stoking demand into these shortages and bottlenecks. And that's creating incredible
01:01excitement. But for investors, we have to be careful to not get too swept up in that and then
01:06look at the second and third order winners down the road to make sure we're balanced and not
01:12getting ahead of our skis on beta in this investment landscape.
01:16I've got questions about the NVIDIA news from last week. $500 billion where six Wall Street firms
01:22go and find third-party capital. And what I wrote about in my column this morning is there
01:26is a difference between depreciation and the economic life of a GPU. Now, as an investor,
01:33how do you model the economic life of a GPU and decide how to raise money against that?
01:38Well, clearly, they're causing an asset class to be built and financed again with that wave
01:42of liquidity.
01:43You do believe that, that it can be an asset class of its own?
01:46I think it can. But one of the things will be the duration of the innovation. We're going to see
01:50incredible new architectures. We know that memory is in an incredible bottleneck right
01:55now. And so I believe that people are racing so hard today, it's hard to re-engineer these
02:02networks. But, Ed, we have to think about three, four, five years from now, we could be re-architecting
02:07and compressing memory and having new structures. What's the durability of these asset classes
02:13the next two to three years? I think it's very potentially in the bag. Five to 10 years,
02:19there's going to be tremendous change. And so I think we have to be a little careful
02:24on the risk and the adjusted return you're going to need to participate in these markets
02:29if you're a longer-term investor.
02:31Alphabet has looked to the equity market and it's looked to the bond market.
02:36And so far, everyone seems pretty sanguine about that. How do you feel about the activity,
02:43but also what you see as being investor demand in response?
02:46Yeah, it shows we're pushing into the more riskier part of this cycle. So initially,
02:51when the cash flow was all out of hyperscalers, you could capitalize that. Now we're going debt,
02:57equity markets. We talk about crowding out. We have a large national deficit and we have
03:03homes to fund for the population. So we're getting into that more risky part of this
03:09infrastructure investment. It doesn't mean there's money to be made. It's just we're at a stage where
03:14it's riskier. In the case study of the $500 billion with NVIDIA and the six U.S. investment
03:21firms, NVIDIA would say there's a degree of separation because it is those six firms channeling
03:27third-party capital, not NVIDIA's own capital. But there is still the circular financing debate.
03:33And for you as the CIO, where does that show up, that concern?
03:37Well, when we look at it, we want to make sure we're thinking ahead of the game and where we
03:42can
03:42participate as investors in the trend without taking on all that leverage and risk. So for us,
03:48in short, it's the second and third order winners. You think about a data center, it starts with a hole
03:53in the ground. So we're investing in Vulcan Materials, which is a toll booth on America's concrete.
03:57We own Lindy, which these purified natural gases, these are 15-year taker pay agreements.
04:04And so when you think about Lindy, they even help send rockets to space with industrial gases that
04:09are very pure, 15-year agreements. And we think about companies like Hubble and GE Vernova, that
04:15these things start with atoms to electrons to energy. So we're thinking about the more 5, 10,
04:2115-year bottlenecks. And that's where I think as an investor, if you want to sleep well at night,
04:26you kind of look ahead where the infrastructure will go. And that's the long-term bankable asset
04:33class, energy, power, electrons, industrial gases, aggregates, materials, those physical parts of AI
04:41are going to be long bottlenecks in our opinion. And we can actually underwrite the durability of
04:45those for a longer period. When you look at NVIDIA, where there is a more direct circular financing
04:52concern is where it is investing its own capital into a project or campus where the
04:59party leasing the compute is a customer of NVIDIA's. But they're still investing in the same
05:06areas that you've just listed. Do you feel good about that or otherwise?
05:11We think Jensen's incredible. We love NVIDIA. We also like companies like AMD,
05:15which has great technology. Their Helios 450 ramp is looking really promising. And they're basically
05:22showing up with a whole range of compute. And they're not as competing with their customers as
05:28much. So that's a company that we think is very well positioned in compute. If you want FPGAs,
05:34CPUs, which are much more important, GPUs, and other embedded options. So that's another way to look at
05:42it where they're not as involved in the circular financing, but has incredible compute.
05:47Todd, you've listed some very interesting names that are being presented with an opportunity in
05:51the build out. What about those names that are most at risk of disruption, particularly in the
05:57software space? Yeah. So Salesforce, Workday, ServiceNow, these are good companies. We all have
06:05friends who work there. These have been great companies. But I think the seat license model is
06:11just going to be under a lot of pressure. You see the token usage. And the minute you see an
06:15anthropic revenue number, that's coming out of somewhere. And you see OpenAI. You're seeing
06:20Google Gemini. So I think the squeezing and the pricing out of that seat model is going to be very
06:26risky. And so we wish them well. They're good companies. I just think the long-term bankability
06:32of that is going to be challenging.
Comments