Netflix’s Warner Bros. Play to Beat YouTube, Ex-OpenAI Head of Sales on Selling AI | Jan 21, 2026

(upbeat music) - Welcome everyone to the information's TITV. My name is Akash Pasricha. It is Wednesday, January 21st. We're gonna kick off the show with Netflix's quarterly results. We will talk about what we learned about the state of the advertising business. We're also talking to A Crew Capital's newest general partner who was coming from OpenAI where she was previously head of sales. And the information is still in Davos. Our editor-in-chief Jessica Lesson sat down with handshake CEO Garrett Lorde and Ventured Capitalist Zach Bogue. We'll play some of those conversations for you shortly. We're also digging deep into context graphs with our friends at Foundation Capital. It is a term that Silicon Valley is buzzing about right now and finally, it is Wednesday, which means we are unpacking this week's finance column, the theme this week. Why price inflation is central to the data center build-out story? It's gonna be a busy show. And so let's get right on into things. Netflix reported its quarterly results last night. The company projected its revenue will decelerate this year to 12 to 14% down from 16% last year. But the company finally gave us a picture of the size of its advertising business to break it all down. I wanna bring on light shed partner Rich Greenfield. Rich, welcome to the show. It's great to have you here. - Thanks for having me. It's been a big and busy last couple of weeks for Netflix. - Busy for Netflix, busy repair amount, busy for Warner Brothers. I wanna talk about all of it with you. Let's start with the quarterly results last night. So $1.5 billion in ads last year. Was that better than you expected? Was it in line? How did you feel about it? - I mean, this is still very early days for Netflix. I mean, they didn't have advertising three years ago. They didn't believe in advertising three years ago. And so, you know, look, it takes time. I think what, you know, what I think we're all looking for right now, if you have an ad account, I don't know whether you are ad-free or ad-cotch, but I have the ads. Okay, so you have the ads. And so what you'll notice is that the ads actually are really boring. It looks like ads on, no, no, no. I mean, there's to be honest, it looks like ads on any other television platform. And I think what the ins, if you think and listen to what they're talking about, whether it's Ted Sarandos or Amy Reinhardt who runs advertising for them, the goal is for advertising to look and feel different. You know, the beauty of Netflix is that it isn't like a traditional programming layout, like where every Tuesday you have this and you watch, you know, from 8 to 11 with a programming grid. And so the advertising can also be different. And I think you're gonna see them start to do more storytelling where as you're binging a series, or, you know, maybe watching, you know, 50 episodes of something, the advertising is actually going to understand that you're binging and rather than seeing the same ad over and over again, you're gonna see a story that's told throughout the advertising. And so I think it takes time for them to, you know, initially it was just, hey, let's just get this stood up. Let's replicate what everyone else does. They have done that at scale working with lots of partners like Amazon, like trade desk for traditional ad business. I think 2026 is where advertising on Netflix begins to look a lot more interesting. And that's why one of the reasons I think you're gonna see advertising double yet again, to be over a $3 billion business. - So it doubles this year, you're saying, my question law, at least. So long term then, because I know you follow the ad space broadly, with all of these platforms, open AI, et cetera, introducing their ads plays. How do you think that then bodes for Netflix's advertising business do you think it eats into it at all at any point? - Remember, I don't think you can really compare story telling, you know, if you're a brand, any brand, and I'm sure information deals with lots of them for your ad business and sponsorship. And you think about like what you're trying to do, you're trying to tell stories. And there really is nothing that tells stories like, you know, watching long form video ads. That's why TV's been as powerful, and especially sports on TV's been as powerful as it has been for the ad industry. If you think about where Netflix is competing, and I would say this is not just Netflix, this is Netflix and Amazon as the real, 'cause both of them have really come into the TV ad space in a big way over the course of the last three years with multi-billion dollar businesses. If you think about where they're eating from, remember, linear TV is still a $65 billion advertising business. That's $65 billion is moving faster and faster over the streaming. And Netflix and Amazon are big beneficiaries. And that's why you're seeing, you know, double digit declines across the linear TV, especially the cable linear TV universe. Those businesses are in a whole lot of pain, and it's gonna get worse and worse in 26. And so, I don't think this is about fighting with meta. This is really, you know, the real opportunity here is to take from linear TV. - Taking share from the linear. - For sure. - For years to come. - So, okay, so if that's your opportunity in ads, then let's shift to the other side of the house, the traditional subscriptions business. And then, you know, I do wanna get your thoughts here on the Warner Brothers deal. I guess maybe if we start there, just give me a thumbs up, thumbs down. You like this deal for Netflix? - This is a pretty bold deal for Netflix. I mean, this is really looking at the opportunity. I mean, look, they are growing subs. Subbs were up about 8% last year, you know, over 325 million subscribers. Certainly not the, you know, 16% growth we saw the year before. And that was just an absolutely explosive year. But I think as you look at growth, they are looking at this. I think, you know, honestly, this is very similar. If you aggregated the three big Disney acquisitions, think about what Disney did under Bob Iger. They bought Lucas, they bought Marvel, Marvel, and they bought Pixar. Those were three big IP acquisitions. I think for Disney, I think for Netflix, this is looking at Warner Brothers and thinking about like, think about the content. You get DC Comics, you get Harry Potter, you get, you know, Game of Thrones. You get so many big franchises and an incredibly deep TV library to mine and to work with. And so I think it's, you know, when you think about the long term for Netflix, and sure the stocks under pressure right now, people are worried that they have to pay more. And, you know, how to, you know, are they doing this because engagement is suffering? But if you think about this over a five year period, this is a huge long term positive for Netflix. If they can pull this off, even if they pay a little bit more than where they are right now, this company is generating $11 billion of free cash flow. They can afford this. And I think they can really do amazing things that they're able to acquire this library of content and this production. And you think it gets through regulators? I do. I mean, when you think about, I mean, the thing that I think most people miss is that Netflix is still smaller. You know, if you, if you look just purely at streaming, sure, Netflix is, you know, if we look at subscription streaming, Netflix is the largest. If you look at all of streaming, obviously YouTube, and YouTube's doing lots of things. You can watch Sunday ticket now on YouTube. You can watch NFL games on YouTube now. But but leaving that aside, the reality is Disney is bigger. You know, if you put all the Disney's assets together, they actually represent more TV time spent than Netflix today. Same with NBC. And so while peacock is smaller while Disney plus is small, the aggregate amount of TV time spent is actually larger than Netflix. And even if you add in HBO to Netflix, you're still smaller than Disney. And so I think from a regulatory standpoint, it's going to be very hard to not look at TV holistically. I mean, you can watch the NFL right on ABC ESPN. You can watch a couple of games on Netflix. I mean, trying to draw distinctions and say the business is only streaming page streaming TV seems like an incredibly narrow definition that I can't imagine standing up in court when they're all fighting over the same programming and the same eyeballs. So then how do you feel about this notion that by making this acquisition, they are clearly getting deeper into this media market that is clearly under pressure broadly. They're also, they're playing with diversification in gaming and video podcasts. I mean, the question I have for you is the opportunity cost of doing this acquisition. I mean, the other approach they could do is go way deeper into diversifying into media verticals that are faster growing and have a little bit more of a tailwind behind them. Do you not think that might not be a better approach to look for growth there? - I mean, look, everything has an opportunity cost and there is absolutely no doubt that you could look at this and say, hey, I mean, why are they not buying Epic games or why are they not buying Roblox? Like you could certainly look at this and say, what are the opportunity costs of things that are being passed or foregone? You know, in terms of investing, I find it, you know, it's, they've been trying to get into games. I think games is not an easy thing to get into for them. You know, they've tried to get into now the more casual TV-based family games. That seems to be working better. They'll obviously have the Warner Brothers Games studio or maybe that creates a launching pad to go deeper into gaming. But I think when you look at what Netflix is, Netflix is bread and butter, right? It is TV and really owning time spent, video time spent, and they're fighting, you know, like the battle they're fighting. - There's a fight on YouTube. - You too, right. - Like you're fighting YouTube, right? And YouTube is getting stronger and stronger. I mean, look at the growth. I mean, you with the information spend a lot of time on the creator economy, Jessica and your team has been early to looking at how the growth of that with conferences and you think about like sort of what's happened there and you look at Netflix there. They're now trying to bring in more and more creators. You know, you look at, you know, they just got Mark Rober or the science guy and like they're bringing in more and more people and doing podcasting, which is sort of a creator economy like business and working more with them. I think Netflix recognizes the threat and how competitive this space is. And they're trying to win long-term and they're trying to fight a company that has, you know, even greater scale in Google YouTube and they see this as a central acquisition to help them accelerate and further, you know, catch up in terms of overall engagement to where Google is. And so I think from that standpoint, sure, there are other things they can do. I don't think this is a break the bank acquisition. I mean, if you think about the size and scale of Netflix, they can digest this. They're at a size where they can do this and it doesn't destroy the balance sheet. It doesn't put them in financial jeopardy. And I think if you look at sort of what they do best, you know, think about this, Amazon. They spent $8 billion to buy the, you know, to buy MGM. James Bond, they had to pay an extra billion to get access to making James Bond's. But when they need to get the James Bond library watched, who do they turn to? Netflix. You can watch all the James Bond films on Netflix because that is what they do best. Netflix is great at really shining a light on catalog and getting it watched. It's why NBC Universal did a deal with them. It's why Sony did a deal with them. And I think that opportunity to increase viewership of all of this content that's sitting, not being watched in that Warner Brothers library is a huge opportunity for Netflix. - Great, well, Rich, I want to thank you for coming on. It was an exciting quarter and I trust there's more excitement to come. That is Rich Greenfield, the co-founder and analyst at Lightshed Partners here on TI-TV. Okay, A Crew Capital has a new general partner. Alisa Rosenthal is joining the venture capital firm from OpenAI where she was previously head of sales. She was also OpenAI's first commercial hire, growing enterprise revenue from $10 million to $10 billion. I want to bring on Alisa to tell us about this next phase of her journey. Alisa, welcome to the show. - Thank you, I'm happy to be here. - Congrats on the new role we've had. Lauren Kaladnyon, the show previously. So we know A Crew Capital well here at TI-TV. What are you gonna be focused on at A Crew? - Yeah, you know, AI has really matured into a platform. And I'm focused on the AI native companies reshaping enterprise through durable infrastructure, developer tools, and applications built on this massive platform shift. And you know, I saw firsthand at OpenAI on the front lines really the gap between model capability and enterprise reality. And I'm excited to invest in the emerging app layer addressing this gap. - Okay, so you're gonna focus on apps. I wanna talk a little bit about your area of expertise, which is sales. That's what you led at OpenAI in large part. The broad question I have for you is how do AI sales differ from traditional enterprise software SaaS sales, which I think is a world that you lived in before AI. So you know both these stories. How is it different selling AI? - Yeah, you know, I think I had the opportunity to build the first really AI native Salesforce. And that meant changing the way we sell. It also means that buyers are changing the way they make purchases. So some examples of this, you know, when you sell ChatGPT and I think when you sell any AI product, you're not just selling a tool. You're selling an entirely new way of working. It's a new technology. People are nervous about it, they're fearful. And you're something of an ambassador of AI. And it was really interesting. I had to tell my sales reps, you know, you're not just selling a tool. You are a thought leader in AI. And you have to be bought in, you have to understand it, you have to use it. And that meant it changed the type of sales person I hired. You know, when I first started at OpenAI, I really looked for people who were excellent at their craft. We'd really been my fostered sales. And by the end of my few years at OpenAI, I was looking for people that had lettered AI to be excellent at their craft. So people who weren't not necessarily as great as the talk, well, they have to be good at the talking, but they also have to be good at using the tools to inform the talking, essentially. Exactly, they have to be good at both. And the more it depth you are at the tooling, and the more you use it in your day-to-day jobs, the more authentic you are at selling it. And the more you can help people who are using it for the first time with real existing apples and real opportunities to make it impact another working. So now on the pitch of all this, we've had some folks on the show who are involved in advising enterprises that are buying AI software companies that are looking to buy sales for service now, Microsoft, et cetera. And one of the comments that they've made on the show here is that they wish that sales teams would have a little more transparency around, hey, you might not get the ROI as quickly as you think. You might need to experiment with this for three, six, nine, 12 months, even, you know, it might not be as immediate as old-school enterprise software in some ways. As you think about that reality and how teams get ROI out of AI, how do you incorporate that messaging or how did you incorporate into sales teams? How do you advise companies to be honest with it and then ultimately convince people to still take a chance on you? - Yeah, so two things I'll say there. The first one is that I think trust and authenticity matters more than it ever has before. You hear people saying sales is going away, sales will be completely automated. And I really think that's not the case because, exactly, we've hit upon, buyers being inundated with AI products, they don't understand what's real, what's not. And at the end of the day, they're making this decision largely based on who they trust and who they find authentic. So I think that's point one. Point two is I think we're at an interesting moment in the evolution of this app layer where it's mostly general-purpose tools that aren't necessarily custom fit for certain types of companies, certain industries, certain workflows. So the solution is to throw forward deployed engineering at it. And that I think is why the integration period is so long right now. You have really a general-purpose agent that you are quote-unquote training on your workflows, on your data. And it takes a while to get that agent to be able to perform the way you want it to. I think in the future in what I'm really excited about is I think we'll start to see more sort of specific applications. And by the way, I use a term application very broadly. I think the form factor of applications might change. They might be background agents. They might live in your operating system. But that aside, I think these agents will have more custom context out of the box. They'll be more specific. They'll be tailored to very specific and unique use cases. And they will work more right out of the box without a ton of customization and tailoring and forward deployed engineering. As you think about all the conversations that you had with customers in your previous role, I'm wondering if there is one conversation that comes to mind that really embodies the challenges of getting enterprises to adopt AI and really what's top of mind for them. You know, what I'll say was very interesting at OpenAI was we learned a lot from our customers. You know, it's most companies you have a problem you set up to solve it and you build a solution. At OpenAI, we did research. The research yielded new capabilities. And then we would figure out how do we productize them and commercialize them, which is a very different way of going to market. And so what this meant is we had to be very close to our customers and understand how they were using our products. So we could package that up in solid accordingly. And one example is we had a tool called code interpreter and it was pretty much using Python. And we thought it was a coding tool. The first customer that started using it very broadly was a private equity fund. And we called them, we said, what the heck are you guys doing with this tool? And they actually came to our office and they showed us how they were using code interpreter to form due diligence or in Black Sholes analysis, due heavy, mathematical, quantitative research in Chachi, BT. So we actually ended up rebranding code interpreter as data analysis and really repackaging it and post-training it in a way that was much better at data. And so that meant we were learning from our buyers and our customers, which I think is a very different mindset for a sales organization and a product organization. Last question for you. We've been talking about ads a lot this week as not just OpenAI, but all these platforms look more towards ads. I know you were mostly involved in the enterprise side of the house. And so maybe slightly different, but still sales is sales at the end of the day. There's a clear opportunity in selling ads and building out that market. What do you think is the biggest hurdle that these sales teams might face in selling these advertising products to retailers and e-commerce players? I think the biggest challenge for OpenAI in this model will be keeping the trust of users. We've talked about trust a lot today. People are naturally skeptical and cynical about this technology. They're nervous about it. So I think for OpenAI to convince advertisers to join the platform or convince the public to engage with these advertisements, they have to be really careful to do it in a way that is subtle, that doesn't feel like the results are being skewed by advertising that their results that are genuine and that the advertisement is related but not changing the results. And that's going to be a hard needle to thread. So what about selling it to the retailers? I mean, the buyers of the ads. What sort of challenges do you think they'll see there? I mean, on the same way, they have to convince retailers that the users are going to engage with the ads and feel bought into the new experience and continue to use the application. Right, great. Well, Alisa, congrats on the new role. We are very excited for you and look forward to having you on more as you make some noteworthy investments over at A Crew Capital. That is Alisa Rosenthal, the new general partner at A Crew Capital here on TI-TV. OK, the information's Jessica Lesson is still on the ground in Davos. In partnership with the World Economic Forum, Jessica spoke with handshake co-founder and CEO Garrett Lord. Handshake has traditionally operated a marketplace for corporate employers to find new talent. But they have now positioned themselves at the center of the AI staffing opportunity where big AI model developers are trying to get their hands on as many human professionals as possible to help them hone their models. Jessica and Garrett discussed that transition the company's making and also where Garrett sees hiring going in the AI era. Here are a few highlights from that conversation. If you think about model progress, like a lot of the gains that have been made so far in very verifiable domains, like answers where you can guarantee get the correct right answer. But if you think about a lot of professional domains, for example, like investment banking or investment making, what is the right answer is very subjective, right? And so you really need human beings to design what are called verifiers or rewards that help models understand what good looks like. And the reason they do that is obviously for this topic called reinforcement learning where if you can teach the model what the right answer is, the model will try many different paths and get to the right answer. And so luckily, I think for humans, like a lot of the world is still very unviroifiable. Like maybe you as the editor and CEO of integration, like what is a good article is very subjective and it depends on human expertise. And so that's what we're very good at doing is hiring tens of thousands of the most qualified people in every discipline and domain. And then bringing them into a very structured environment, teaching them about AI, how to break the models, and how to design data that helps the frontier climb. So as you're pushing into this area, others are to scale, got a number of big headlines when it's founder decamped to beta but is still a business that operates in this space. What do you see in terms of competition and how can handshake compete? So obviously, you know, we have many amazing competitors in the space and Kudos to the entire, you know, many of our competitors that operate today, where we really see our lean is in this kind of decade of institutional trust that we've built with 22 million people. And so we're able to, you know, our core business was recruiting, building trust, acquiring people, helping them find jobs, helping them learn information for jobs. And so on top of this kind of decade of trust across 1600 universities in the country, we've really been able to leverage that to turn around more quickly than our competition. The labs, I think at its core, they care about three things. They care about like the speed of which you can turn around. They care about the quality of the data, really quality is paramount. Like if you're, you're really powering the data that's helping models go climb. So quality matters a ton. Speed quality and then volume. They'd be able to get to heavy, heavy volumes. They don't care about costs. They do care about costs. Okay, we just left that one out, but yeah. Goss, I think it's a good part of life. Yeah. And our network really provides us like this incredible strategic advantage of being able to turn around more quickly, with the high volumes than our competition. And that's why in eight months, we've climbed up, you know, high hundreds of millions of dollars of revenue and the business, it feels like we're in early innings of what it could become. If you think about like LOMs, it's really, you know, you have compute, you have algorithms, and then you have data. And I think on the compute side, compute scaling in a huge way, algorithms, you continue to have a ton of algorithmic process. And I think in the data side, which, you know, is really our business, like we're still in, we're in a world where like the entire internet's basically been scraped. Like Andre Kaparthy talks about, you know, kind of like exhausting the fossil fuel of what's already been on the internet. And there is just still so much intelligence and human knowledge that's still out there, but it isn't in like some Reddit form with like a million posts that I can learn from. And so on the data side, if we can continue to kind of translate the frontier human knowledge into data that helps models climb, I think as compute scales and algorithms scales, models will continue to get better. And I really dream of a world where like, you know, each one of our employees in handshake is like immensely more productive at work, and able to spend their time doing the things that they want to be doing. And less time like figuring out how to build like a looker dashboard for six hours. - Yeah, okay, no more dashboards. I love it. - Jessica also spoke with Zach Boag, co-founder and managing partner of venture capital firm DCVC. They discuss investment opportunities in nuclear fission, data centers in space and biotech here are few highlights from that interview. - Many people are calling new advanced, new advanced nuclear fission. So this is fission. This is, you know, they're similar to the existing nuclear technologies, which currently provides about 20% of the power in the US. But we have credible groups that are claiming they will be bringing new advanced nuclear fission. So these are new types of nuclear fission online and as soon as 2028. And we have a portfolio company called Radiant Nuclear, which says Micro Nuclear Reactor. So these are one megawatt, think about powering a remote army base or a remote community or-- - This isn't the backyard ones. I've read about those two. - These aren't the back of them. One megawatt is still a lot. But it's not, you know, it's not going to, you know, power a hyperscaler database or data center. They are credibly going to turn on their first live reactor this year sometime in the Idaho National Labs dome test facility. So this is a dome where you can turn on reactors in a low consequence way and get the data and use that for, you know, the subsequent approval. When they turn that on, that'll be the first new advanced nuclear reactor turned on in the US ever. - Wow. Have space data centers crossed your portfolio yet or thinking about that new frontier we're just starting to hear about? - So yeah, DCVC, we've been investing in space for a long time in space tech. Obviously, Planet Labs was a really bed of ours from, you know, over a decade ago. - I ran into Will, the CEO, right out there yesterday. - Yeah, exactly. - Will is terrific. And we have looked at some of the space data centers. You know, our orientation having invested in a lot of space companies. The space is still very expensive to get stuff off the planet. Obviously with SpaceX, there's been, you know, plunging launch costs, which is crazy. Just across the course of my career, I mean, it's over a 10X decrease, you know, in a per kilo basis for getting into space. I still wonder if it's going to be cost competitive to launch a whole bunch of, you know, Nvidia chips into space, which are, you know, it's heavy and cooling and expensive. And every once in a while, those rockets blow off. But it's exciting, but I still, you know, I think, you know, we really need to understand the unit economics of why it makes sense to do a compute in space. So we're big investors in many of the leading tech bio companies. And I'll define tech bio. So tech bio, in our view, is a company that starts out first and foremost to build a novel, highly structured proprietary database, which is ready made to be able to run AI algorithms on or other machine learning. And then they use that to go off and hit targets in therapeutics or diagnostics. So it's sort of the reverse of a typical biotech company. That's why they've sort of reversed the term. And these companies also, it's a little bracing. They take longer than sort of the traditional biotech company where you're coming in and they already have sort of some development candidates identified. And it's really building this computational piece first that they then go off and do that. So that's what we're excited about right now. We think we're in the early innings of tech bio. - Presumably AI helps there. I think if you're sifting through data, it seems like a good AI use case. - Absolutely. And these companies are predicated on the use of AI. This is kind of an AI play because they're building these data sets that are ready made for AI. And this is companies like Recursion Pharmaceuticals, Relation Therapeutics in the UK, and Noetic, which is a company of ours that just had a, we think for the first time sold for $50 million, sold access to just a foundation model in biotech, which is we think is the first of a kind deal, which is pretty exciting to see just sort of bits changing hands, not adubs in biotech. - Yeah. - Okay, one of the latest hot terms in AI right now is context craft. It is a fancy name for describing a number of systems and softwares that interact with each other to enable AI and agents. Our next guests are two of the first people to really help bring attention to this idea. Jaya Gupta is a partner at Foundation Capital and Ashu Garg is general partner at the firm. Welcome to you both. - Thank you for having us. So look, I'm gonna be honest, I tried. Okay, context graphs. I mean, it's new, it's a little dense. - Jaya, maybe you were the one that wrote the blog post and maybe I'll give you the first track of this. Can you explain it to me as if I was a person that was not in tech? Let's start there. - Absolutely. So if you take a step back, 2025 was supposed to be the year of AI agents. And I think the models did get better. So ChatGPT, Cloud, all the models, they got better. But in enterprises, they still don't act like you'd expect them to. And I think the reason is actually super, super simple. Agents can read data and take action, but they don't actually know why the decisions got made. And that reasoning is what we call decision traces. They're scattered across different tools. They're buried in Slack. They sometimes don't even get reported at all. And so we believe that the winners will be the companies that capture those decision traces and turn them in to context graphs. - So the way that I'm thinking about this and I'm asking here, and I'm not declaring that this is the way it is. But you've got the agents, you've got the systems of record, which I think is a fancy way of saying, this is where the data is tapped. Is the right way to think about a context graph, the software that connects the agents to all these databases in a more efficient way and allows them to pull on data better? Is that the idea, Jaya? - I think it's more than that. I think of context graphs very simply put as the institutional memory of what happens, why it happens, when it happens. And the reason this matters is that if you look at the last generation of enduring software companies, they were built in systems of record that captured the outcome. They captured what happened ultimately. The next generation of software companies we believe will be built on context graphs that capture why decisions are made and the turtle through which they're made. The reasoning behind the decisions, which is mostly in the heads of human beings today. And we believe this is a trillion dollar of our creativity. - And so is this a new class of software that's gonna be created that we're sort of saying there are gonna be new players in this space? Is that the idea, Ashu? - But we believe that this is a new foundational layer for AI software. In a few years, all companies are using the same models. The question we have to ask ourselves is, what is the moat for software companies? And we believe that moat is the context graph because it is the institutional memory of how decisions got made and not what the final outcome was, which is what we start, please, been captured in system director. If it's useful, I can give you an example. - Yes, please, we like examples. - So imagine for a minute, you're out of Salesforce. And the Salesforce app says, hey, we always give healthcare companies an extra 10% discount because there's procurement cycles of brutal. That information that incites of an extra 10% discount for healthcare companies, it's not in the system of record, it's not in your CRM, it's institutional knowledge. In another situation, a sales VP might approve a 20% discount to get a deal done on the same day. That approval happens on a Zoom call. The systems of record will show the final outcome, the actual price that it was sold at, not who approved it, why the exception was granted or what precedent was used. But let agents try to automate this the next time around because that's the whole idea that agents will automate that renewal next quarter, they have the outcome data, but not their reasoning. And that reasoning is really the missing layer, the decision trace, and the data of those decision traces become the conference. - Okay, so Jaya, so it's coming into focus, so this is the software that is going to basically help the agents basically behave more like humans in terms of helping to interpret the subjectivity of all this decision making. Jaya, my question for you is, we just heard the highlights from our conversation with the CEO of Handshake, and they belong to a class of companies that are involved in helping the labs deal with complicated queries and decisions and really turbocharging reinforcement learning. It really involves having humans train a lot of these models better, and to my understanding, add that nuance that Ashi was talking about. So how does that story then swear with the context graph segment that you're talking about? - It's a great question, and I think that there will actually be many, many different players that in the space that we'll be able to sort of benefit from the context graph category. And I think one of them will absolutely be companies like Handshake, companies like Turing in that space that this is also going to be a new class of data. And at the same time, where we see this going is that every single application company that will be built will have to fundamentally think about how to incorporate this architecture. And we will also see many, many different infrastructure companies get started as well here. And we're already starting to get pitches, which is super, super exciting. And, you know, if you guys are building in this space, like we'd love to learn more. - And Ashi, are there companies that are doing this right now? - Absolutely. And you know, the thing I would add to our data set is the most unique context graphs are the ones that are built as white companies that are in the workflow and decisions happen. You kind of capture these decision traces after a fact. And so we have, I'll give you an example of a company. I have a portfolio company where Boji and I worked very closely, called to Sarah. And the founder, Kabir Nagrecha is a 21-year-old PhD. We met him before he had a team, before he had a product. And his mission was to automate SAP migrations. And we kept asking him what becomes defensible when your competitors have access to SAP models. And it was that series of conversations that led us to our thesis that our decision creates in context graphs. - And, but our companies, I'm thinking of companies like Gleene and at last year, are they doing this? - So actually not. - In order to have a context graph, you have to be in the workflow of these decisions. We believe these systems of agents like to Sarah, Maximor, and many others in our portfolio, Plano Zero is doing this for software that actually capture these decisions as they happen. You know, what we call in the right path as the decisions flow through the organization, those are the companies that are the best position today. Now, that doesn't mean other companies can't get there. And Gleene's an amazing company with a lot of stress. But they're not actually in the right path today. - And, Jaya, one question I had for you is we've talked about the corporate data wars on this show and systems of record companies adding walls to being able to access the data in a world where, and look, I think people come on the show and they've said the customer will inevitably decide and the walls will inevitably go away over time. But that does seem like a hurdle to this class of startups getting traction at least at the outset, right? - So I think it's right, but I think as you rightly pointed out and I bet other people in the show is that, you know, the customer is at the end of the day or the ones who can speak up and who are already starting to speak up and can say, hey, like we need access to, you know, we need access to this thing for this sort of application. And so, yeah, so I think that at the end of the day, like the customers have the power here. Plus, it's getting easier to access this data. I mean, in the old days, you're gonna have APIs, now you have APIs, but frankly, you can reverse engineer any API using the same models. You can also reverse engineer the front end of an application. So the idea that there'll be a wall garden around systems of record, I think is a myth. - And Ashu, last question for you here, does the business model here, does it look different from the way that we've seen enterprise software companies move from subscription-race pricing to consumption-race pricing, usage-race pricing? What ultimately is the business model for this class of startups that you think will be successful? - The common theme around this class of startups is they're automating services, services that are done typically by white color individuals. You know, and this is an insight that we sort of published and I heard a couple of years ago, this idea of services as software. And in services as software startups, the business model ultimately will end up being, can you capture a share of the value you create through that automation? It could get price in the form of a price per rating, it could get price in the form of a price per unit of output. So we think the business model will change because you want to share in the value. The key though is in order to share in the value that you create for customers, you have to build a defensible mode and that defensible mode for system, for services software companies will be your context for. - I said last question, but I just have one more actually and Jay, maybe I'll end with you. What is the technical challenge here? If this is the idea for the class of startups, what needs to get better? Do the models need to get better? - It's a good question. I think the technical challenge will also be, like it will actually be the opportunity for most startups and I think the bottleneck for enterprise AI is not actually going to be intelligence. I think there's a lot of talk about doing work for centuries but I can't think about the last time I did a task that took more than, that wasn't more than a week. And so I think context is gonna be the biggest bottleneck and I think what's missing is going to be the decision traces, what inputs were considered, which exceptions are granted and what precedents really applied. And I think context graphs capture reasoning and I think traditional systems or records are actually never designed to store those decision races as well. - Great, well, Ashu and Jay, I wanna thank you for coming on and explaining it all to us. That is Ashu Garg and Jay Agupta from Foundation Capital here on TI TV. - Thank you. - Okay, building data centers at the pace at which big hyper scalers and AI companies are going nowadays comes with a number of challenges, not the least of which is coping with costs that lately have been rising quickly. My colleague Ken Brown wrote a column in today's finance newsletter about these bills and why price inflation is so central to this story, I wanna bring them on to talk all about it. Ken, welcome back to the show, it's great to have you here. - Hi, Akash. - Happy Wednesday, let's talk finance, okay. - Yeah. - So which costs are going up here exactly in the data center story? And you can't say everywhere, everything. - A lot, a lot, I won't say everything, but a lot of costs. The cost of electricity is going up, power, cost of some the gear that you buy to run a data center. Simple stuff like copper has gone up a lot and AI data centers use a lot of copper, a lot more copper than regular data centers. So it's just, it's a lot of different things. Okay, and how are companies getting around this? I mean, copper, you know, are they hedging? Are they innovating, what's the solution here? - So right now, and for the foreseeable future, there's not much you can do. Or if you can manage one, you're stuck with another. So it is just, the demand has been very high for this stuff. There's been underlying inflation in some of these commodities and it's all sort of come together. There's a lot of things going on that could change things in the future, small nuclear reactors and better management of power, better more efficient chips, all that stuff. But that's all in the future. And as we know, you need the computing power right now today. And so that's the stress going on in these developers. - So the thrust of your column was really looking at what the net impact will be on these data center development projects then. And the obvious hurdle here is getting over these costs. I mean, what do you see then playing out here? Do these projects slow down? Do some of them get shut down? What are the repercussions here? - Right, so look, nothing is stopping the AR and Juggernaut. So there's still going to be getting built. And the people who are funding or the companies that are funding these, Google and Meta and Microsoft are cash rich and they can handle it. So that's going to keep going. The question for them is going to be at a certain cost. Do I keep rushing? Do I slow down? How do I think about that? The pressure is going to be on the smaller players. The some of the data center developers, some of the companies like Oracle or CoreWeave who are in the middle of all this. And they just don't have that kind of money. And so they're dependent on raising money. And so when investors are looking at these and they're saying, well, I don't want to lend you this much money for this much data center. I don't think we're going to get a return on this because the cost is too high. That's the sort of thing that's going to go on. And we're seeing mentions of it. We're seeing a lot of discussion of it. We haven't seen things not happen or projects get canceled. My view is that's going to start to happen this year. And they're going to be the projects on the margin. And you may not even hear about them because there's some project that someone's going to do and they pencil it out and it just doesn't work and they don't do it. Is there at all a cascade that could happen here in terms of, you see a couple projects get canceled. And then, and I'm just trying to think of the ecosystem of players where you talked about the way that they get financing here. And there's obviously the debt component to it all. I mean, I guess my question is, is each project really considered an isolation in terms of whether or not it actually goes through or is there a bit of a cascade risk here at all? And I'm asking if someone who hasn't really followed real estate, which I know is so central to this story. - Yeah, I mean, I don't see, I don't see a cascade. I mean, I think there could be companies that run into a bit of trouble, right? They may not hit their growth targets. But that doesn't mean they're gonna be like, disappear. They're just gonna be, you know, not as robust and not as attractive as they were. I think, you know, the issue is gonna be, does this slow the pace down? Is there, you know, are there any, you know, big events, you know, for example, in the last, well, in the last few days until today, interest rates have been going up a bunch because Trump has been screaming about Greenland and stuff like that. Well, like, if interest rates go up much more, that's gonna be an issue, that's gonna be an issue across the board. And so like that could really grind things down. So we just have to see how all this plays out. It's, it's, I think the individual companies is gonna be depending on their financial situation, whether they go ahead with this stuff. And underlying that is how much money they think they could make from AI. So like we saw over the weekend, open AI released some data on its growth. And essentially, it was almost a perfect match, which was every year, for the last three years, computing power has tripled and so has revenue. So computing power equal, computing power growth equals revenue growth. And so if you don't have the computing power growth, that's gonna slow you down. - Right, right. And we've certainly seen companies coming out and saying, I think it was Andy Jesse actually, who said to, who told Jessica in Davos, he said, if we had the capacity, I mean, we would be growing a lot quicker. So that certainly seems to be the consensus right now. Ken, I wanna thank you for coming on the show. It's, it's certainly a story that we will keep monitoring and look forward to having you back again soon when there is more news to talk about. That is Ken Brown, our finance editor here at the information. Okay, well that does it for today's show. A reminder, we are on this stream Monday through Friday at 10 a.m. Pacific, 1 p.m. Eastern. I wanna thank you all for tuning in. We really do appreciate your viewership. I'm already excited for our next show tomorrow. Have a great rest of your Wednesday. Bye bye for now.