Podcast: Why AI’s Venture Opportunity May Be Era Defining

Key Takeaways

• AI may drive a $7 trillion to $10 trillion economic expansion over the next decade, potentially adding 10% to global GDP, according to various market predictions.
• The cost of training and running advanced AI models has decreased exponentially. The rapid decline, coupled with the increasing speed of AI advancements, is accelerating the adoption and impact of AI across industries.
• AI is likely to lead to profound shifts in labor markets, productivity, and the structure of multiple industries. For example, a majority of coding could soon be performed by AI models, while AI likely is having far-reaching impacts on working practices in healthcare, legal and business services sectors, to name a few.
• On the risk side, AI investors continue to address challenges related to resource intensity (including capital, energy, and hardware) as well as questions around data security, societal impacts, among others. Another common question revolves around the positioning of incumbents versus startups. While incumbents hold advantages including capital availability and data, nimble AI-native startups that are focused on applications and tooling could build significant share.
• Leading venture firms will continue to attract top founders by leveraging their brand strength and experience. But alongside established firms, AI-native venture funds are emerging that have been started by investors deeply embedded in the AI ecosystem, making them attractive to founders. 
   

 

 

Transcript

Marcus Lindroos: Hello, and welcome to the Adams Street podcast. I’m Marcus Lindroos, a principal on the primary investment team. We’re joined today by two Adam Street partners, Brijesh Jeevarathnam, the head of our primary investment team, and Stephen Bluestein, also from the primary investment team.

We’re going to focus our conversation on artificial intelligence and hopefully provide some perspective on why Adams Street believes that AI presents a compelling opportunity for venture investments.

It has been a dynamic last five years for venture capital, with a period of significant activity during 2020 and ‘21, a period of normalization in 2022 and ‘23, and a pickup of activity in ‘24 and beyond, with AI being a key driver of interest and activity in the last couple of years.

On that note, Brijesh, let’s start with you. There are a lot of headlines about AI today. How would you simply describe the state of AI and where you’re seeing value being created?

Brijesh Jeevarathnam: Thank you, Marcus.

I would say we are at a unique moment in terms of where AI is driving a fundamental shift in how we do everything from learning, creating, working, and living our lives.

I think it’s important to take a step back and think about the Gen AI technology wave is still relatively young. While we can trace back to machine learning and different predecessors to Gen AI, the technology we’re seeing now is really maybe three to four years old, and we’ve come a long way in a very short time.

And what AI is doing today to answer your question is, I’ll break it down by a few different areas, starting with the technology side and then maybe quickly the impact side.

On the technology side, we’re obviously seeing a few themes play out in front of our eyes.

Number one, the cost of developing this breakthrough technology is coming down at a pace that is unprecedented.

One of our managers, Elad Gil, talks about, for example, the cost of training inference for a GPT-4 level model has come down 180 times, not percent, but 180 times in the last 18 months. Just think about that for a second.

That is a massive component of what’s happening here. And along with the cost decreasing at that level, the speed at which newer iterations of Gen AI are coming out will keep progressively getting faster.

And if you combine those two factors – lower cost, higher speed – the advances and therefore the impact that Gen AI applied technologies, which we’ll talk about more in a couple of minutes, could be really massive at a societal level.

I heard the CEO of Entropic, one of the AI Gen models, talking yesterday. And he was making a prediction that within six months, 90%, 90% of all coding will be done by models and within a year, potentially 100%. Now just think about that. Software and coding, obviously, is a large part of our economy and a large part of our tech economy for sure, and even if the prediction is off by a year, two, three, four years, it doesn’t matter. That’s the scale of change that we are talking about, potentially, driven by Gen AI.

That’s on the technology side and I’ll just go back to the economy side for a second. If you take some of the macro predictions, Goldman Sachs, for example, predicts $7 trillion of economic expansion in the next 10 years from AI.

And the CEO of Microsoft, Satya Nadella, was asking aloud and really kind of  postulating that AI could increase global GDP by 10% in the next 10 plus years. We’re talking about, once again, around the world across all parts of the economy, game-changing disruption and innovation coming from Gen AI.

Marcus: Pretty incredible changes we’re seeing in the last couple of years. You know, Stephen, maybe pivoting over to you. There’s a lot of discussion around where that value is going to accrue that Brijesh just talked about, whether it’s going to be in incumbents or startups. Where do you see startups as being best positioned to generate value versus the tech incumbents today?

Stephen Bluestein: Thanks, Marcus. I think this is one of the most seminal questions that we, as allocators and investors, are thinking through in this moment, because we’ve had this massive wave of technology that Brijesh talked about come to fruition. And now we’re thinking about where is their value going to be captured. I think let’s try to break it down in a couple of different dimensions.

The first dimension is along the hardware dimension, sort of the infrastructure dimension. And what we saw, starting back a long time ago, was this cloud wave and the build out of AWS, the build out of Azure and Google Cloud.

And what’s so compelling about some of those vendors and what the position they have in the marketplace today, is that they are in a real catbird seat to be able to take advantage of AI. Most of the models need to run, obviously, in the cloud, and they’re going to run on this hardware and this infrastructure that’s been put in place by these large incumbents.

And so on the hardware layer and the infrastructure layer, it’s very natural to believe that the current players in that market – Microsoft, Google, and Amazon – have a real structural advantage against the potential startups that are out there.

Now the big question around the hardware space is will [there] be proprietary models or open source models. And so, we are seeing a handful of new interesting hardware providers. You’ve got a company like CoreWeave that’s going to potentially go public pretty soon. You’ve seen other hardware providers like Together AI, which is stitching together different models together.

And so you are seeing derivatives that open source may become the dominant model, and if that happens, potentially in the hardware layer, well certainly, Google, Microsoft, and Amazon will take advantage of that as more infrastructure runs on those clouds. But they may not be the de facto infrastructure winner. We may see new winners emerge over the course of the next cycle. It’s still TBD. It’s still very early as we’ve talked about, but we are seeing new infrastructure models.

And then, of course, part of infrastructure is tooling and how to run these models. And that’s where you have a lot of legacy software providers that have gone public over the last cycle. They’re obviously adapting their tooling to work in a Gen AI way. But I think we believe pretty strongly that there’s going to be a next generation of infrastructure companies, that startups are really well positioned that are AI native to take advantage of what’s happening in the AI space at this moment.

I think infrastructure is probably where you’re going to see a lot of the incumbents do really well. But at the same time, you’re going to see a handful of new tooling companies that pop up.

As we go through the stack, the area where we’re probably most interested in seeing a lot of … where we  think a lot of new innovation is happening, where a lot of startups are going to capture value is specifically within the application layer.

In there, obviously, it’s very easy to believe that a lot of the legacy application providers that are out there in the market today, that have been bought by large companies, can refactor their applications and make them AI native. I think that’s going to happen for some legacy application software providers. I think the ones that can do that, obviously, are going to be really well positioned because they have distribution, right. They already have the customer.

And I think that’s a huge advantage, a moat that those players – think about Salesforce. Salesforce has hundreds of thousands of customers today. If they’re able to adapt their technology to AI, why would a customer rip out Salesforce? But the flip side is this technology is so new, and whether or not Salesforce has the people and the ability to adapt these solutions for the AI native world is an open question.

Where we’re going to see a lot of AI powered applications that are native to AI, we think it’s going to be oftentimes in legacy industries. Think about healthcare. There isn’t a dominant healthcare software provider today. There’s obviously things like Epic software that is run by hospitals, but if you look at the fragmentation in the healthcare space, thousands of applications, very few with a real dominant market share position.

So that’s an industry where we think a lot of innovation is going to be happening on the AI native side. It’s an area where data is so critical to unlock the value and no one software solution provider in the last generation of software was able to capture that value because they weren’t able to stitch all this together.

Right now we’re excited to specifically in vertical focused applications. Think about legal. Another area where I think, lots of data on people’s drives, hard to put that data necessarily in the cloud for legal and protection reasons. And so we think that the ability to use AI to unlock that data in industries that have been very hard to do are very natural where we think new startups will emerge.

Again, it’s going to be a battle. People obviously have different moats, but we’re excited about the application layers. We think there’s lots of innovation that’s happening there.

Marcus: Amazing. I think a really helpful framework to think about the investment opportunity set here. Brijesh, maybe we can dig a little deeper there and we can talk about a few examples where we’re seeing real tangible value being driven by companies in our venture capital portfolio.

Brijesh: Yes, Marcus, and I think Stephen covered really well.

If you don’t mind, I’ll just add one thing to what he said, which is this AI, Gen AI wave, is one of the few where the incumbents aren’t fully at a disadvantage. Stephen said this, and the way I think about it is the incumbents – Microsoft, Google, Meta, et cetera – they have at least three distinct advantages potentially over the startups.

There would be obviously data, massive amounts of data. Just think about how much data Meta has or Google has by owning YouTube, I mean the parent company Alphabet, et cetera. That is important for Gen AI as we’ve talked about.

The second is resources including capital, and some of these investments are massively resource intensive, and the incumbents have the advantage there.

And the third, as Stephen said, is distribution. You can already touch customers, and even with a product that’s less good than the best startup out there, you can capture market share.

I think the startups here have also potentially three advantages relative to the incumbents.

One is, of course, the agility of just being a small team, and we see that all the time in the technology space. A team of 20 can do something that a team of, let’s say, 2000 can struggle with, and innovation, agility, all of that.

Secondly, that goes with that territory, is the quality of talent is potentially higher on the startup side. One of the proof points for that is the fact that we’ve seen a few “acqui hires” essentially by big companies of these smaller startups in the last 12 months.

And the third advantage potentially is, unlike a legacy or an incumbent player, which has to start with the legacy set of modules, code, architecture, this is clean sheet start, which gives them an advantage.

But, unlike in other prior waves, where the incumbents really were fully disadvantaged, here is more of an almost a quote unquote level playing field and depending on what we’re talking about, both the startups and the incumbents, they work closely together. OpenAI is a good example, working with Microsoft because that’s the way to really get to the end game faster here.

I think that’s a good segue to your question, Marcus, about applications that we’re seeing, and really there’s quite a wide range and in the interest of time, I maybe pick one or two verticals. I’ll actually start with cybersecurity.

It obviously is a large and growing area with the surface area for attack for all of us as we use and access the internet across multiple devices, as we store more and more things off premise, i.e. on the cloud, the surface area for attack, and the modalities of attack, are increasing, not decreasing, and therefore this is a more and more robust, attractive area of investment for venture.

We’re seeing a lot of AI adoption by both, I would say, established startups, quote unquote, but also the new startups coming out. We have companies in our portfolio that really have taken their offerings to the next level in terms of adopting AI and going to the customer with an even more robust product and making their products more sticky. That’s an area where we have a lot of exposure. I’m starting with that.

Another area is healthcare. Obviously, this is a massive part of the US GDP, 20% approximately, and it’s a massive part of the global GDP, $9 trillion plus global industry. This is an industry where there’s a lot of room for innovation, room for better outcomes for the patients and cost reduction as well.

We’re seeing a number of companies in this space going off to multiple areas. So one example is diagnostics and even prevention, which help with radiology, reading, and more accurate summaries and prognoses. We’ve had companies that help with earlier and more comprehensive detection of cancer.

And when you think about patient-doctor, patient-provider interactions, there’s a lot of room for improvement. Stephen touched on this before, both in terms of the systems used, like Epic, but also in the way the interaction happens and all the information is captured.

We have a number of AI-powered, really much more than transcription tools, they’re really AI-powered note-taking tools that can take the key takeaways from a conversation and help the physicians not just become more efficient with their time, but also help connect the dots with respect to both the diagnosis and potentially the treatment.

But if I go quickly into a lightning round, if you will, we have companies today that are providing real ROI around, for example, legal, , that really reduce the cost and time to do the research needed to represent clients in terms of looking at precedent judgments, in terms of  looking at case law and presenting a cogent strategy for the lawyers to work with the clients, etc.

We’re seeing another area broadly I would call, and I’ll stop here with this category, is business services. That’s a very broad definition. I put that in a few different buckets, maybe at least two. One is customer support and we just heard about a company recently, which is helping the legacy industry companies. Just think about a company that might be distributing food products or something that’s very non-technology, in other words. They also use technology, but those technology systems are very old, written with code, and in languages, perhaps, that are not even in vogue anymore. The portfolio company of ours helps update that legacy code to more state-of-the-art language and coding.

and that’s a massive ROI for the end customer here because they don’t have to deal with 1990s technology anymore. So I’ll pause there because I can keep going with so many examples of real impact in ROI from applied AI.

Stephen: And maybe I’ll just add one example as well, because I think it’s the base layer of kind of everything, is software development. I think this is a really interesting discussion point, because there are 26 million software developers today, and w hat is being built today is limited by the productivity and the number of software developers there are.

As a result of what’s been happening, and Brijesh mentioned this, that potentially north of 50% of the code that’s being generated at companies today is being generated by Gen AI software development tools.

I saw a note today that Garry Tan at Y Combinator said that 70% of all of the code that’s being generated by YC companies, some of the most innovative companies in the world, was generated by code software and not by actual coders themselves. That is just a phenomenal step, because what you have there is a recognition and acknowledgement that the software development process I s being refactored.

That just tells you that the rapid pace of innovation that’s happening in software development is phenomenal and that obviously is what is so exciting is that there’s so many new potential applications that are going to get built, solving real world problems. But it starts with the fact that code generation is happening within software, less with people. This makes the whole process more efficient.

Marcus: Pretty amazing case studies in terms of the breadth and width of the impact across the economy. I think we can all agree that, if we can improve patient outcomes and reduce costs in the healthcare system, that would just be a win, but it feels like it’s touching every part of the economy.

I think, as we talked about this being a $10 trillion net new opportunity, there’s likely room for both incumbents and startups. We just talked about many inflecting very early and creating real impact.

We’ve talked a lot about the opportunity. I think it’s massive, some great tailwinds. But Stephen, maybe we can talk a little bit about some of the risks with AI that we’re seeing and really how startups and VCs are addressing these issues.

Stephen: Yes, I’ll take a first step at that. I think one of the things that Brijesh mentioned was just the capital intensity of building out this infrastructure.

I think that is obviously a risk in the sense that the amount of capital that will go into this and the potential payback is at this moment because it’s so early and its development is a little bit unknown.

And so when we talk to venture capitalists, we talk to startups, the constant thing we’re trying to probe and understand is exactly how much revenue is being generated by these applications. We just gave you some positive examples. But the amount of capital that’s been required in order to build some of these tools and build this infrastructure is really large. You’re seeing just like open AI raise gobs and gobs of money and the rationale for that is they’re building out training and inference. This is a continual process to make these models better.

And so t he amount of capital that’s going into this wave, there’s obviously some analogies to the telco build out back in the 2000 timeframe in how long that took. Again, we’re big believers that this is a massive transformation, but the timeline for the payback on that capital is one of the risks that we’re thinking about.

That has implications for companies that are more infrastructure and hardware related, but it really also has a lot of impact on the startup ecosystem as it relates to what is the gross margin that some of these startups are going to be able to charge for those businesses?

We were having this discussion yesterday with one of our managers, and the hypothesis was potentially very positive, in the sense that the amount of revenue they can charge for these applications because they’re so transformative would be higher than the last generation, but the gross margins may be lower.

Obviously, these model providers have a lot of moat around the technology that they build. And so there’s definitely a lot of risks with it as it relates to startups, because they have to use this infrastructure – they don’t own it themselves. And so there are a lot of risks whether or not with what’s the long-term steady state gross margin of some of these businesses are. That’s a risk that a lot of our managers and a lot of the startup ecosystem is thinking about.

There’s obviously lots of other risks – data security and data privacy and data accuracy. We talked about how AI enables us to unlock a lot of data that’s been trapped on different applications in different places and provide outcomes that are potentially really beneficial to different industries. But the quality of this data, how it’s being used, how it’s being processed, that is still a risk and it’s still something that a lot of companies today are dealing with.

That’s always going to be a risk. It was a risk obviously in the last generation of startups, but I think this risk is even more important here because we’re using all this data to make predictions. And so the quality and accuracy of that data is super valuable to building these models.

Then there’s obviously this risk that there’s one model to solve all problems. I think there’s this idea that OpenAI is going to build everything for every application, or Microsoft or Azure will. I think that’s a risk that could potentially come true because these model providers have so much of the data and they’ve aggregated this data and provided the model. Could they move up the stack and could they be able to produce the applications that are going to be the next generation of AI native applications? We think that obviously there’ll be some elements that these model providers and infrastructure providers will be able to provide applications, but we’re still really bullish on individual AI native applications. That’s a tension in this moment.

If you look back to the last cycle, Amazon, Microsoft, Google all really were very infrastructure focused, Microsoft being the most heavily application focused given its Office suite. There were still tons of applications that were built that were not part of those three different suites. But maybe the model providers today have learned from the last generation and they may move quicker to build the application layer on top of their infrastructure. So that’s a constant tension that we’re thinking through from a risk perspective and how our venture capital managers are navigating that.

Marcus:  Super helpful. We’ve talked a lot about the opportunities here. We just covered the risks. Given all that we’ve discussed, how would you say Adams Street is approaching the AI opportunity set from a manager selection perspective?

Stephen: I think some things here continue to be true and continue to persist that I think is core to how Adams Street thinks about investing in managers.

This is a huge technology wave, very similar to technology waves in the past. We truly believe that the best venture capital managers will partner with the top tier founders building innovative and disruptive companies that can generate long-term sustainable value.

One of the things that’s true about venture capital that we believe is that those brands persist over long periods of time. The Andreessen Horowitz’s, the Benchmark Capitals, the Sequoia Capitals. They obviously have been successful over the last generation of technology innovation. We think many of the strongest brands in the venture capital ecosystem will continue to be able to attract the best founders to want to work with them.

And we’re very fortunate at Adams Street that we have a really long history in the venture capital ecosystem – started in the 1970s and we’ve been partners with many of the leading venture capital firms over long periods of time. These firms tend to be very much generalist. What’s really positive about that is, these firms are able to adapt and learn from the past cycles and be able to adapt to the new innovation that’s happening in the marketplace today. We’re seeing that today. A lot of managers are diving into AI, figuring out what’s happening in the AI ecosystem, and they’re able to adapt to the founders. The key piece is that they have resonance with founders. When founders decide who they want to partner with on their cap table on this journey of building the next great companies, they’re going to look to the brands that have been around for a long period of time and we do believe that the power law is still very much in play here, probably even larger maybe than in the past. And so it’s super valuable to capture that outlier of returns. We do think that the larger, well-established brands have a great market position to capture that innovation.

But to be clear, with every innovation cycle, and this happened during the cloud revolution as well, new firms will emerge that are AI native, much like the AI native companies that we’re so excited about that are thinking about this natively, that have been in this ecosystem for long periods of time, that have the relationship with founders, that are there early, where founders will look to and say, “Well, they’ve been in the AI ecosystem for quite some time, and they understand what it takes to build AI native companies, and we want to partner with them.”

You can open up the papers, you see firms like Nat Freeman and Dan Gross called Century Fund, you’ve got Sarah Guo at Conviction, a whole host of firms today are being started around this theme in this area, and we think that there’ll be innovation within the venture landscape.

And so I think for Adams Street, we’re constantly looking and finding new potential managers that are out there that are very deeply AI native. But again, really excited about the persistence and the strong brands that we’re partnered with, but always constantly looking for new people who are tackling these new problems with innovative approaches.

Marcus: Amazing. Well, I feel like we could talk about this for another two hours easily. But I think that’s all we have time for today, so I hope you found our discussion about the opportunity in artificial intelligence useful. It only remains for me to say thank you for joining us, and we look forward to listening into the next edition of the Adams Street podcast. Brijesh, Stephen, thanks again for your time.

Stephen: Thank you.

Brijesh: Thank you.

Important Considerations: This information (the “Presentation”) is provided for educational purposes only and is not investment advice or an offer or sale of any security or investment product or investment advice. Offerings are made only pursuant to a private offering memorandum containing important information. Statements in this Presentation are made as of the date of this Presentation unless stated otherwise, and there is no implication that the information contained herein is correct as of any time subsequent to such date. All information has been obtained from sources believed to be reliable and current, but accuracy cannot be guaranteed. References herein to specific sectors, general partners, companies, or investments are not to be considered a recommendation or solicitation for any such sector, general partner, company, or investment. This Presentation is not intended to be relied upon as investment advice as the investment situation of individuals is highly dependent on circumstances, which necessarily differ and are subject to change. The contents herein are not to be construed as legal, business, or tax advice, and individuals should consult their own attorney, business advisor, and tax advisor as to legal, business, and tax advice. Past performance is not a guarantee of future results and there can be no guarantee against a loss, including a complete loss, of capital. Certain information contained herein constitutes “forward-looking statements” that may be identified by the use of forward-looking terminology such as “may,” “will,” “should,” “expect,” “anticipate,” “estimate,” “intend,” “continue,” or “believe” or the negatives thereof or other variations thereon or comparable terminology. Any forward-looking statements included herein are based on Adams Street’s current opinions, assumptions, expectations, beliefs, intentions, estimates or strategies regarding future events, are subject to risks and uncertainties, and are provided for informational purposes only. Actual and future results and trends could differ materially, positively or negatively, from those described or contemplated in such forward-looking statements. Moreover, actual events are difficult to project and often depend upon factors that are beyond the control of Adams Street. No forward-looking statements contained herein constitute a guarantee, promise, projection, forecast or prediction of, or representation as to, the future and actual events may differ materially. Adams Street neither (i) assumes responsibility for the accuracy or completeness of any forward-looking statements, nor (ii) undertakes any obligation to update or revise any forward-looking statements for any reason after the date hereof. Also, general economic factors, which are not predictable, can have a material impact on the reliability of projections or forward-looking statements. Adams Street Partners, LLC is a US investment adviser governed by applicable US laws, which differ from laws in other jurisdictions.

This podcast is a recording of extemporaneous remarks delivered in an informal, conversational format. Statements herein regarding characteristics of Adams Street Partners, its investment process, past, current and potential investments, and market conditions or forecasts, as well as other statements are necessarily high-level in nature and should not be relied upon as a basis for any decision, including an investment decision. Many such statements are forward-looking statements and are associated with numerous known and unknown risks, uncertainties, assumptions and limitations. Before relying on this podcast as part of any decision, including an investment decision, a listener must contact Adams Street Partners to determine whether there are additional relevant material facts and/or contextual information that are necessary to understand the statements herein, as well as whether there are any risks or limitations associated with the information described herein. All such information will be provided upon request. All characterizations or comparisons are the subjective impression of the speaker and need additional context in order to be reasonably understood (which could not be provided contemporaneously due to the informal nature of the conversation). In particular, statements of differentiation, such as describing advantages, favorable comparisons, quartiles, deal quality and other statements should not be relied upon unless the same statement appears in written format in a document prepared by Adams Street Partners for its investors.