The Information's TITV
00:49:36 · Amanda Peterson Corio, Stephanie Palazzolo, Nicole Su
[00:49:36] S7: I am much more worried about the risk of underbuilding than overbuilding. We have very clear demand from our customers, and we have a variety of customers. Of course, we have our standard Google products and search and YouTube. We have our, Google Cloud and enterprise customers through cloud in addition to Gemini. So we're fairly confident that, everything that we're building today and paying for is going to be used and utilized. [00:50:04] S0: Are you ever worried though about... And maybe it sounds very out there, but the world where Anthropic or OpenAI is right, and they gobble up all these AI apps and all these customers and startups and companies that are using AI models. Maybe they just get eaten up by the big labs. Like, I feel like there's so many crazy scenarios out there of things that could happen. How do you, you know, what happens if one of those far out cases does come true, and who's left holding the bill? [00:50:33] S7: I think in every scenario that you're mulling through, it still needs infrastructure to operate. Whether that's power infrastructure or data center itself, is an asset in and of itself that inherently has value. And so when you're creating an asset the right way, that has value, there will be a use case and a need for it, no matter what unknowns play out and how this all evolves in the [00:51:01] S3: AI space. Well, we'll have to come back in two to three years and
00:51:01 · Nicole Su, Stephanie Palazzolo
[00:51:01] S3: AI space. Well, we'll have to come back in two to three years and [00:51:04] S0: see see whether we we under or or overbuilt in that case. But it was really great to have you on the show. Thank you so much, Amanda. And again, that was Amanda Peterson Corio, Google's global head of data center energy here on TITV. Nicole Su leads AI investment banking in The Americas at UBS, advising companies navigating one of the biggest investment booms in tech history. We're going to talk AI investment, valuations, and IPOs. Thanks for joining us today. Welcome to TITV. So, you know, as somebody who's covering AI, I think one of the biggest trends I've noticed is every time it feels like we're reaching a new peak of valuations, it just goes higher. You know, so what are you seeing now? Are investors starting to become more disciplined at all, or is FOMO continuing to drive prices higher? Maybe taking
00:52:04 · Nicole Su
[00:52:04] S3: a step back, I think we're definitely seeing a lot of sky high valuation headlines, but I wouldn't necessarily translating valuation headlines into the actual cost base of the investors. I think you have reported it on yourself as well. There's definitely a multi tranche structure that's happening right now. At the same time, like there's... Also could be terms in the term sheet that would actually bring down the cost for the investor as their cost base. But back to your question earlier, I do think FOMO and being disciplined are happening simultaneously. Think there is a small group of companies or talents people believe could create generational companies. So that's where scarcity comes from and where valuations keep getting higher. But at the same time, I think investors are getting more disciplined because I'm seeing investors are definitely start thinking about, in addition to the top line growth, right, what's the cost of the growth? What's the sustainability of the business and how long until this company can start making money. I think all these questions are being asked these days when
00:53:07 · Stephanie Palazzolo, Nicole Su, Brittany Greenfield, Jessica Lessin
[00:53:07] S0: it comes to diligence process for investors. And you mentioned two things. You mentioned these kind of like multiple tranche rounds that we've written about where, you know, maybe the headline number is 3,000,000,000, but actually investors got in at 1,000,000,000 and 2,000,000,000 and 3,000,000,000. And you also mentioned maybe there are some other terms that are more investor friendly. What are some examples of these like other terms that are, you know, maybe more friendly to investor that you're seeing? [00:53:29] S3: Maybe the the most common example will be warrants. I think a lot of them will come with pending warrants. I think that would be a good concrete example to give. I think others is too bespoke to really talk about. [00:53:41] S4: Mhmm. Mhmm. [00:53:41] S0: And are are these signs that power is shifting over at all to investors? Or do you think the power still lies in the hands of the founders for these hot rounds? [00:53:50] S3: It still totally depends. I think quality of the investors, quality of a team, right, the scarcity of the assets, it all depends. [00:53:59] S0: And you mentioned that there's both FOMO, but also investors are trying to be more disciplined. Does it feel like any part of the market right now is especially overvalued or certain parts of the industry? Like, what do you think is the most overvalued these days? Tough question. [00:54:16] S1: You can't say everybody is.
00:54:16 · Jessica Lessin, Nicole Su
[00:54:16] S1: You can't say everybody is. [00:54:18] S3: But but at the same time, I do think from my seat, right, I'm seeing up and down the AI stack, but also vertical apps as well. I think what I'm seeing is I'm seeing a lot of companies, particularly in the application layer, they have tremendous growth. But if you look under the hood, right, how much they own when it comes to intelligence and how do actually have a flywheel of intelligence as they accumulate more customer data. And then what's really the sustainable business model and a sustainable moat that can defend, right, as frontier models also getting more specialized? I think that's a big question mark. So I think I'm excited to see, like, how these vertical apps companies that would, you know, really come up on the curve and come to getting a sustainable business model can drive profitability alongside of growth. Mhmm. And that's one. And then the other one is everybody's favorite topic these days is physical AI. I think it's really because the promise that, you know, bringing generalized AI into the physical world is huge. The TAM is much bigger than what we're doing in this digital AI world. But I think it's also, again, it's the difference between the expectation of how big the TAM is and the pace it's going to take to actually unlock that TAM. I think we're still in very early days when it comes to building general purpose robotics models. A lot of promising progress, but I think we haven't reached the tragedy moment anytime soon.
00:55:44 · Stephanie Palazzolo, Nicole Su
[00:55:44] S0: And I feel like, you know, speaking of physical AI and the world of, you know, robotics and all that craziness, I feel like there is a bigger debate now between, you know, like there have been some studies that have shown that actually the general language models are actually pretty good at maybe directing robots better than people expected. I guess I'm curious, do you feel like there's still a strong argument for, you know, robotic specific models versus everything just being controlled by, you know, general language models? [00:56:16] S3: I think the reasoning part, I think, is totally transferable. I think it's more about interaction with the physical world and especially when it comes to specific form factors, because I'm seeing really good progress when it comes to robotics model architecture evolution, right? We started with the BLM and VLA and our world models allow people to predict the future, but at the same time, like we are missing previously, like inputs like tactile data, but I've seen some of these leading research teams have started publishing paper about how they actually were able to integrate all this tactile data into their foundational model that has all the elements of LLM for reasoning, L VLA for, like, interaction between vision and action, but also word model for prediction. So I'm seeing... I think it's gonna be a combination of all these model architecture to work together to unlock the physical AI world. Yeah. So kind of like