B (34:09)
We've been talking all day, all week about the concerns around the amount of debt that firms are taking on in support of the data center buildout. So there's another D that could throw a wrench in the boom depreciation. Bloomberg's Dana Bass, he covers AI infrastructure, has been writing about the lifespan of AIG PUZ in Bloomberg Businessweek, you join us now. Michael Burry has been talking about it, other players have been talking about it. Should we have anxiety about depreciation of GPUs? Dina? So it's very hard to say. We're going to talk a little bit, Carol, about accounting one on one here. So when you buy a lot of physical assets, a company has to decide what the useful lifespan of those assets is and write it down over the course of that time period is called depreciation. GPUs, which companies like Metta, like Google, like Microsoft, like Open Air, spending tens of billions of dollars on them. We really only have a couple of years of useful data for how long the current generations of those will last. And so most of the companies are still say, are writing them down over the course of five to six years. But there's a real concern because as we all know, Nvidia is committed to putting out new ones annually and is sort of trying to obsolete its own product. So there's a real concern about what happens to those, how long can they be used? If you're writing them down over too long a period of time, you might be artificially boosting your, your profits. What was interesting was in Nvidia earnings, they said, look, our A100 hundreds, which have actually been officially discontinued, they're still up and running, they're still working and efficiently. So they were trying to sort of put some calm amid this anxiety. You've got a great quote coming from Sarah Fryer, The CFO of OpenAI, of course, who'd been talking about how they're structuring their GPU depreciation. Sure. So Sarah Fryer said to us, look, you know, on the one hand we don't really know is it four years, is it six years? But what OpenAI has seen is that they are, they feel good that it's at least five. And the reason is that they know that they're still using their A1 hundreds. And the basic idea for, you know, companies like OpenAI is use the latest chips, the Nvidia Blackwells, for things like the very high end training of the absolute top of the line frontier models. The older chips can still be very useful for things like inference. So running the actual models. And so that's the way that they think that this all works. But that all requires, you know, the data centers to be what everyone's been calling Satya Nadella has been calling fungible. So it means you can switch what the data center does from, from training to inference. You can switch from one customer to the next. So when a customer rolls off and a chip is a little older, can you find a new customer, take that older chip? That's a lot of the question around, you know, quote unquote, how fungible these data centers are. Well, core Weave was fungible from being a bitcoin Miner with its GPUs to them being an AI company. And many of these neo clouds have pivoted in that way. Dean and lastly we're looking at, I looked at KKR for example. They're talking about potential froth in AI investment more broadly and they're sort of trying to understand how the end user is using their data centers. But, but really who is on the hook for all of it? They're trying to say, look, datacenter owners as operators, often it's going to be the leaser who is guaranteed long term payment. Do we know who really has to hold the baby here and the bath? I keep asking people the same question, you know, when the music stops, who doesn't have a chair? Because as you mentioned about, about debt, there's been, you know, large amounts of debt contracts, you know, financing of these GPUs and if the GPUs are useful for less time than we think, what are the people that, that you know, are holding debt on them? Do? Do they ask for more collateral? Do they, you know, how does this all go? And you know, people that we spoke to did say that regulators and investors have to think about how far the quote contagion is going to spread if this all, you know, sort of comes home to roost at once. Dana Bass, on the optimistic side, we appreciate it. Lovely to have you in town amid this Thanksgiving holiday. We appreciate it. Meanwhile, markets look they're trying to weigh the risk of depreciation on one side, but there's also the hope, the hope that I will pay off, particularly say in life sciences and health care discoveries. Already major players like Anthropic have stepped into this area with models aimed at boosting research and development. Now investors see an opportunity to. Lily Lyman is managing partner of Underscore VC is a Boston based early stage investor in startup startups like Tetra Science, H1 Quilt Health here in New York, not Boston. You really think Boston though is going to potentially be a winning trade when you think about the confluence of what's going to win out if health care is supercharged by AI. Well Caroline, thanks so much for having me here in the studio. It's great to be back and look, as investors are always looking for a clear why now in a market opportunity. And we are certainly seeing a signs of transformation in life sciences and AI. What's changed is that we now have the data, we have the sophistication of the models, and we have real industry pull to unlock this potential. And what it could do is unlock $4 trillion worth of value across life sciences and health care. What's changed is that biology is no longer just a wet lab discipline. It is now a data and information industry. Many people are calling it tech bio Mm. And we had underscore as pre seed and seed investors, and particularly based in Boston, are incredibly excited about this because it's opened up a whole new world of software investing opportunities in the world of science. Can I go to that 4 trillion number? Yes. What's that pegged upon? What is it that we're seeing that will be fueled and to garner $4 trillion of worth? Well, if you think about what the life science, the combination of life sciences and actually, honestly, all of scientific data and health care, it's across the gamut of how this gets done today. I mean, Today it takes 2 to 3 billion dollars per drug and decades to develop these therapies. Think about if that can get cost, you know, if that can get changed to be a fraction of the cost in a fraction of the time. The economic and the human impact of that is absolutely enormous. So the opportunity we see is across the entire value chain in life sciences. So whether it's research and discovery, whether it's in preclinical and clinical trials and services, manufacturing, development and deployment, you put all that together and it's not hard to see how there could be a $4 trillion opportunity coming out of this. And you said you're seeing signs. What are the signs and what are the software companies that are leveraging those signs? We're seeing across science, across all the different players in the market. So the major pharma companies are certainly making moves in this space. They are under enormous pressures. I mean, the cost of R and D is rising. It basically doubles levels every nine years. They are facing issues with their margins. They're facing a potential $260 billion revenue cliff as some of their patents expired. And so that's creating the market pull for AI solutions. They are partnering with often many times startup companies. So, for example, Takeda just launched their partnership with Tetra Science, which is a company we're invested in. Think of Tetra Science like the snowflake, but specifically for scientific data and what it it's doing is it's partnering with all the major technology players, Nvidia, Google, Microsoft, Databricks, Snowflake and rallying the tech stack around this opportunity to unlock scientific data so it can actually be used by models. And so in this one what they're able to do is working with Takeda on hundreds of use cases so I can sit on top and use this data. And it's driving 90% faster workflows, 40% increase in productivity. So we're seeing that type of adoption and partnership across the major pharma players and startups. How does a portfolio company compete with an anthropic who's getting into a similar space? People always love to ask that question, how do, how do, how do startups compete with the incumbents? And I always think that yes incumbents have the advantage of data and distribution, but startups have the advantage of focus and speed. And so we're seeing those opportunities across the board. I mean there's couple a company we're investing in called Terraflow which is automating the analysis and data around flow cytometry. And again I mentioned Tetra Science. You know, these are opportunities that require very specific domain focus, very specific types of people who can do it and the ability to build in an AI native way from, from the ground up. The anthropics of the world. Open Air is also launched in this space. They also are going to need to do the practical implementation and so they're going to need partners along the way to do it. So I think it's not a zero sum game. I think there's an opportunity for collaboration. So do you think so. Lily Lyman is over in New York for a short while. We appreciate her coming into the studio. Managing partner@_ VC. Online travel and experience booking company Peak. It is doubling down on Air. It's acquiring Acme Ticketing and Connect and Go. And the move positions Peak to expand its reach across museums, theme parks, tours and other attractions. The company also raised additional $17 million in funding. Here to talk it all through its Peak CEO, Rizwana Bashet. So you call yourself the shopify of experiences. Explain what that is. Basically with the operating system that works with museums, tours and activities providers and we provide all of the tooling they need to run their business. So online booking and payments, everything that you do on site is you're checking in all the way through to marketing, business analytics, collecting reviews. We really are the end to end backbone for everything that a business needs to operate and that backbone is solidified by your M and A. So talk to us a little bit. What Acme brings to the equation. What Connect and go, how are they fueling the growth? Absolutely. So you know, what we saw over the last couple of years, we really got to know these businesses and they've done incredible jobs involved working for very specific verticals. And so as an example, you know, Acme has built incredible infrastructure and ticketing for museums and iconic cultural attractions. Think about that as here in New York, things like the MoMA, the Whitney Museum, the Frick Collection. So that includes memberships and donation management. And so they've done a fantastic job there. That's something that we can incorporate into everything that we're doing at peak and connect and go. Really doubled down on theme parks and on water parks. And so that means that RFID technology that you've probably used when you've taken your kids in a water park this weekend. For my sons. Exactly. So you've used that. So they've done a great job on, on, on that technology alongside a lot of things around online. Sorry online and, and on site guest services, things like fmb. And so, you know, bringing these three companies together, we get a huge advantage by being able to have a lot of synergies as well as being able to take all of the best features and cross pollinate them across the platforms. And the last piece obviously is just that we've, we've been real innovators on the AI side and so we're now in a position to take all of the things that we've learned and take them across. So what sort of innovations in A.I. really? You know, last year we went and polled our businesses and over 80% of them said we know we really want to use AI, but only 10% were using AI. And so it became very clear that for us to be able to assist those merchants, we actually needed to integrate AI tooling into our platform. So examples of that have been first on revenue growth, which is obviously incredibly important to businesses. We, we created a dynamic pricing tools. So that means we're incorporating things like weather or seasonality or frankly local cycle demand. You know, think Taylor Swift is coming to town. And what we were able to do with that was layer that into the pricing for the businesses to increase revenues by about 5 to 20%. So massive impact on revenue growth. Another area that we've really done a lot on AI has been around automating operations. As you can imagine, the businesses we work with, they have a lot of manual, you know, backend operational tasks, you know, Bryant park ice skating here in New York, very popular this time of year. They have lots of people trying to reschedule, so they're spending thousands of hours on. On these manual tasks. So we were able to automate all of that work and in doing so, save them thousands of hours of time as well as millions in costs. And I think the last thing has really just been that there's a shifting consumer demand landscape. We all know that over the last few years, people have been moving towards things like TikTok for the video content. You know, over half of consumers say that they're inspired to book experiences based on what influencers think. And yet the businesses we partnered with didn't have a way to be able to meet that demand. And so we created influencer marketing tools. With a click of a button, they were able to reach hundreds of thousands of potential travelers flows. And so in doing that, what we're really allowing our operators to do is, is focus on what they're really good at, which is delivering an incredible customer experience while taking care of what is a huge shift in the industry. But that shift comes with costly talent. You've just raised $70 million. Is that what that's for? Is about beefing up your own tech talent to be able to bring more generative AI offerings to bear. Is it about more acquisitions? Where does that funding get put to work? Yeah, it's absolutely. It's about us consolidating all of the. All of the platforms as well as really layering in more AI. So think about, you know, the things we've already done to automate operations. You know, we've now got hundreds of agents working behind the scenes 24, 7 to do everything that the merchant needs. And so, you know, what we're really doing is doubling down on innovation. And so that means tech talent. And it means also, you know, an opportunity. Opportunity for us to double down on sales. You know, one of the things that we saw with the acquisitions is that although Acme and Connecting Go have fantastic customers, they've actually done very little on the sales side. So we want to bring those tools to market. So, you know, AI plus Sales allows us really to get our tools into the hands of many more businesses and a few more experiences for all of us out there. This Thanksgiving and holiday season. Rizwana Bashir coming on, talking about the M and A and the fundraiser peak. We appreciate it and that does it for this edition of Bloomberg Tech. Do not forget to check out our podcast. Find it on the terminal as well as online on Apple Spotify and Iheart. From New York, this is Bloomberg Tech.
