AI in Venture Capital: Separating Hype from Opportunity

Artificial intelligence is reshaping how institutions invest, operate, and fulfill their missions, and the rules of responsible engagement in practice and in portfolios are still being written. Welcome to Espresso Chats. I'm Amanda Novello, senior policy and research analyst at Common Fund Institute. My co host, George Settles, is out in the world and will catch up with us next episode. But I'm here to share the news that today we're kicking off a mini series on AI. Over the next few episodes, we'll be delivering short, strong shots of insight from leading voices at the intersection of AI markets and institutional governance. We'll be exploring what it means to lead through this moment to balance opportunity with real and evolving risk, and we'll be asking what it looks like for institutions to get it right in this moment. Today's conversation zooms in on the pocket of the market where tech companies typically get their start venture capital. We've also discussed the infrastructure and energy side of the story in our January episode on real assets and AI as a private equity theme in our private equity episode a few months ago. So check those out as well. But right now, I'm excited to welcome to the show, my colleague, Luke Rossiter, Managing Director on Common Fund's venture capital team. Thanks for joining me today, Luke. Thank you for having me on, Amanda. Very much look forward to the conversation. Perfect. Can you start by sharing a bit about your work? Everyone's talking about AI, but you're in it on the day to day. So what is your team up to? What are you seeing in this space and what's exciting to you about it? Sure, yeah. So from our perspective, the impact of AI is very real. For those listeners not aware, our role is to invest in technology venture capital funds and then in many cases into startups directly also. Over different periods, there have been waves of technology innovation that when they occur lead to huge value creation captured by startup companies. Think mobile, think cloud computing, think software as a service. And AI is primed to be the largest wave we've ever seen. The scale and speed at which the industry is progressing and how companies are demonstrating measurable return on investment is unprecedented. The speed of disruption we have seen since the dawn of generative AI and the large language models has been so quick that traditional incumbents are having trouble keeping up. When this happens, it's a unique opportunity for startups and innovative companies to take market share. Those are companies that venture investors are putting into business. Much of this value is being captured in private markets by venture capital investors today, and examples could be large language models like Claude from Anthropic or ChatGPT from OpenAI, or even application layer companies like Harvey in the legal sector. But importantly, it's still really early and we're only beginning to see AI's true impact on how we live and work. And you asked also what is exciting. I think really the most exciting thing about AI platform shift is what it does to the total addressable market or the TAM for software. It's already clear that the market opportunity for AI is multiples larger than what we've seen with traditional software. Historically, traditional software companies sold primarily into technology budgets, which is just one line of a company's overall spending. That meant that the total market was limited to what businesses were willing to allocate in their budgets to software and IT. AI companies, however, are breaking out of that box. Instead of simply offering tools that help workers become, you know, fifty percent more productive, AI systems are beginning to automate entire tasks and in some cases these technologies don't just augment human labor, they're actually starting to perform it. Now what that does is that opens up entirely new markets and use cases that traditional software could never have addressed, and as a result, AI companies aren't just competing for software spend anymore, they're going after the much larger pool of labor spend, the money that companies typically spend on people rather than on tools. As you alluded to, AI is really dominating headlines, but headlines aren't investment theses. So, out the hype from the real opportunity is maybe one of the most important jobs for investors right now. So, can you help ground us in where we actually are in the cycle that you referred to, and what are the real opportunities at play? Yeah, sure, Amanda. I think that would be really helpful. Perhaps I'll start by stating the lay of the land. Artificial intelligence is not new. It's in fact been around for multiple decades, but we have seen it progress rapidly over the last ten years or so. So think ten years ago, AI could not really tell the difference between a dog and a muffin, and now it's scoring ninetieth percentile on exams, so huge leaps in progress in the last decade and particularly in the last kind of few years. When we look at the progression of AI in the 2000s it was all about predictive, so think about Google search. In twenty ten, it was more about computer vision, think things like facial recognition, early iterations of autonomous cars being possible, and then the more recent breakthrough has been generative AI. The breakthrough was not just the chat GPT, last language model, but rather the breakthrough was when the tech community realized the potential of generative AI. So despite the Transformers paper that kind of founded generative AI in twenty seventeen, it hasn't been until the last couple of years that people really understood for the first time computers could be creative in a way not possible before. Ultimately, the goal was to move then from kind of those prediction identification and then to general AI, which today can provide co pilot creativity reasoning capabilities. Longer term, I think the goal will be full autopilot, so to speak, And to explain that perhaps it's reductive but today's models are good for Q and A but tomorrow's will take actions on your behalf such as execute bank transfers or call people on your behalf. Agents will transform agents will transform into co workers where work is done for you and alongside you, and I think this is where the real opportunity lies and what we get very excited about. When our managers are evaluating AI companies to determine the hype from the opportunity, they're really asking a handful of questions, call it five questions. What's the moat or what's the sustainable competitive advantage for a company? So if OpenAI or Anthropic ship their next model release and your product becomes a feature, you maybe don't have a company. We want to see proprietary data sets, workflow lock in, or distribution that's really hard for the models to replicate. Second, are people actually using your product? AI gets massive curiosity traffic in a way that we haven't really seen before in software. That can churn very quickly, so people trying new AI systems and then realizing they're maybe not for them or isn't what they want to spend their money on. So here we look for revenue retention, increased usage over time, not just pure number of sign ups. Thirdly, do the unit economics work? Every query does cost real compute and real money, so we want a credible path to healthy gross margins, not we'll figure it out later when we're at scale. Fourthly, are customers paying real money for real workflows? At the moment, we would put less emphasis on pilot programs. As I mentioned, there is this level of curiosity that we're seeing, but you know, does that transform into real revenue? So we want to see expansion of contracts, multi year contracts, and really evidence of replacing budgets, be that software headcount or services with AI services and software. And fifth, does the team fit the bet? Vertical AI needs founders who deeply understand the workflow, infrastructure plays need elite research and technical talent, and so having the wrong team for the wrong opportunity can kill a lot of these companies early. Perhaps if I was to kind of summarize that into a simpler way to think about it would be, does this company get stronger or weaker as foundation models get better? If better models make them more valuable, that's a real business. Thank you so much. We on this podcast love a lot of detail. We also love a summary. So you covered all the bases. Thank you. Clearly, there are opportunities and those again were helpful questions to parse that out. But I have to ask about the flip side, there are economic and social risks around displacing workers, environmental energy and water stress is only getting worse. And then on the investment front, there's this astronomical capital expenditure with a somewhat uncertain or yet to be determined return on investment. And last but not least, Common Fund has written about concentration risk, where a majority of venture dollars are going into just a handful of companies. And that's not to mention the public equities story as well, which could be another episode. So for institutional investors with an endowment or a foundation, how should they think about this tension between excitement and risk in a portfolio? Yeah, it's a great question, and venture has and is inherently a risky asset class. Balancing excitement about an unknown future has always been part of the equation. Essentially, there would be no reward without the risk. However, the investors that do well tend to be the ones who spend just as much, if not more time, thinking about what can go right as compared to what can go wrong. In the venture business, it's often pessimists who get to look smart in the short term, but it's usually the optimists who end up making money over the longer term. So rather than just ask the question about how AI will disrupt or displace workers, why not ask the question about what new jobs will be created thanks to AI that didn't exist before? Or if AI is making each person in the workforce more productive and let's just use twice as productive as a benchmark, wouldn't you want to hire even more workers at that increased productivity level? So that's some of the positive thinking. However, you can't ignore flip side of the argument for AI, and investors are right to have a skeptical eye. Valuations are increasing, a lot of money is flowing into the space, and you have to be honest about the real uncertainties while still recognizing we're in a completely different paradigm shift to prior technology shifts. When I think back to the internet, there are a lot of companies that raised money, were high profile, and ultimately no longer exist today, but it did also produce Google, Amazon, Oracle. To your point on concentration, venture has always been concentrated by design. We often talk about the power law concept in venture where the top one percent of companies drive the bulk of returns across the industry and how it s critical to concentrate your dollars in those companies in order to generate strong returns. For institutional investor, the goal isn't necessarily to manage perfect diversification in the asset class like you may in other asset classes. That's actually by definition a recipe for average returns. It's to make sure you're concentrating your capital in the right names, the right venture managers that get you into the right small handful of category defining names that are founded each cycle that go on to generate the lion's share of exit returns in the industry. Again, going back to the internet days, think the Googles, the Amazons, the Oracles, If you had exposure to those, your portfolio did very well. For those optimistic but discerning investors that we're talking about here, and speaking of manager selection, if the winners in the AI race are being founded and backed by venture managers today, is it the case that manager selection is perhaps as important as ever? With so much capital flooding into AI focused funds, how do you evaluate which VC managers have an edge versus those that are just trying to get on board for the ride? And for institutional investors who are maybe one or two steps removed from the company level, you know, investing through managers in this way, what are the most important questions they should be asking to ensure they're investing in companies with real potential? Yeah, this is an interesting conversation and really a conversation about alpha versus beta. Riding the wave may get you average returns. Typically, there's a large gap between top quartile returns in venture capital and average or median returns in venture. So it's important to be intentional with your manager selection if you want to outperform those benchmarks. In periods of rapid change like we're in today, I would argue selection of managers matters even more than ever actually. The capital flooded in makes selection harder, not easier. AI focused fund is often a marketing label today versus a real thesis when you receive a manager's deck, and it's important to understand that. The honest cut here is there are managers who had an AI thesis before twenty twenty three and managers who repositioned into one. And I have a point of view as to which of those are best positioned to succeed, but I think it's important to be able to differentiate that, which comes through experience and being deeply embedded in marketplace. Some of the things we look at to evaluate which managers actually have an edge: sourcing, not just selection. So we ask which specific deals they led or anchored and which they passed on and why. Riding the wave managers tend to show up later in rounds at higher valuations. The real edge here is showing up as proprietary access at the early stage. We look at the technical depth on the team. Who in the partnership can actually evaluate model architecture, training data, and inference economics. If the answer is we use advisors for that, that's probably a flag in this cycle as well. And then another thing we look at is a specific thesis. So we invest in AI isn't a thesis, we push for where in AI, where in the stack do you think value will accrue in this cycle? Will it be the foundation models? Will it be the infrastructure layer? Will it be the application layer and why? We've talked before that venture is an access class, not just an asset class, meaning it really can work for you if you have access to the best managers and perhaps isn't worth pursuing if you don't have access to those very strong managers. We see performance dispersion as wider than any other asset class, and unlike most strategies, past performance in venture is actually quite predictive. That's to say that top performing managers show the ability to replicate performance across vintages and cycles. The better their performance, the stronger their brand, stronger their network, the better their deal flow for new entrepreneurs starting companies, more likely that those entrepreneurs want to work with them and so they can win competitive deals. So there's a total flywheel effect. That makes access to proven, cycle tested names the central question, and those funds are heavily oversubscribed with strong preference for existing LPs. So I do believe that, on the whole, established managers have an edge in this cycle. We've been investing in venture since nineteen eighty nine, so that's over thirty five years of consistent strategy and relationships. Many of the names founders are choosing today to partner with are the same names we've been partnered with over decades. Alongside that, we also back emerging funds, who we do think have a chance to be enduring franchises of the next few decades. But as the flight to quality is intensifying in artificial intelligence, we see the best founders want deep pocketed partners who can help them, scale from idea to IPO, and so we see that they're broadly going to more established managers with larger funds, not new entrants riding the wave. Although I would argue there are, selectively a handful of more emerging managers that, whose strategy can be successful in this environment. Luke, thank you for all this context. Thank you for helping kick off this series as we help listeners parse out the hype from the real opportunities and help listeners