1Mby1M Virtual Accelerator AI Investor Forum: With Rajeev Madhavan, Clear Ventures (Part 1)

Rajeev Madhavan, Founder and General Partner at Clear Ventures, discusses how he is investing in AI startups and what he is learning from the market.

Sramana Mitra: Today, we have the pleasure of Rajeev Madhavan’s company here. He is the Founder and General partner at Clear Ventures. Rajeev has had a long and successful career as an entrepreneur before that. He’s a very hands-on guy that I’ve known for a very long time. I actually did a lot of work in the electronic design automation world, which was Rajeev’s old world. Rajeev founded a company called Magma Design Automation, which was a very courageous thing to do at the time because EDA at the time was an oligopoly of Cadence, Synopsis, Mentor, and Magma.

So anyway, Rajeev has deep experience, and I invited him back to the program here to brainstorm about today’s contemporary cutting-edge topic, which is artificial intelligence.

As you heard, for those of you who were here last week, there’s a lot of confusion in the market, including among the VCs. This is why we started this new discussion series to get a little bit of clarity on what exactly is happening in terms of AI investments and how do we parse the trends, how our investors think about things and how should entrepreneurs think about things. Rajeev, welcome to the show. It’s great to have you back.

Rajeev Madhavan: Thank you for having me here, Sramana.

Sramana Mitra: So Rajeev, what is your investment thesis in the age of AI? How are you looking at the world and how are you thinking about investments in this world?

Rajeev Madhavan: So, I’m going to kind of divide it into a broadly two or three categories. But before I do that, I just wanted to give a line on what Clear Ventures is because how much money we have will determine what we invest in, right? We can only put things into what we can afford. You can’t just act like a $10 billion venture fund when you’re a $330 million venture fund, right?

We are a $330 million venture fund here in Silicon Valley. We look at seed and series A – it’s very early stage, we don’t need the first customers. We most likely would have funded a company where there’s no PowerPoint even of what they’re going to be doing. It’s like sitting down with them. There’s a rough idea of what it is. And that’s where we invest in.

Obviously, AI investment started a lot earlier before the generative AI boom, where everybody can become an AI expert. Suddenly, that’s what’s happened in the market, but the AI boom has been divided into simply three things, right?

One is the high-cost chips and semiconductors that are designed using AI for AI in many cases, sometimes just for AI. You’d mentioned my background as being in semiconductors. We met with every one of them because of 99% of them were my customers’ friends in my past life. I chose not to invest in that because my biggest customer was Nvidia. And I don’t think taking on Nvidia on inferencing now, which is what the latest trend is from chip design, is that easy given the software framework that needs to happen.

Because of all the semiconductor restrictions, there’s a lot of money pouring in from Middle East funds, and Chinese funds are trying to get in. So, it’s a large set of investment swath of money that is going into that area. I tend to believe that in 2026, maybe one or two which will get aggregated into one of the cloud suppliers, but it’s not going to be the best of exits in that space given the large number of players that have come in and the large amount of money that’s gone into that space.

The second layer on top of that is what I call the application builder layer, right? It is the basic core – AI ops or ML ops; it’s called by various names by various VCs. This is the layer where you have players like OpenAI providing generic large models and Hugging Face providing a repository of models. You’ve various companies in that space. It’s a very costly space because you end up paying the fees to Nvidia in that space. If you’re a company in that space, you end up spending a ton of money on training, on model building, and on making sure that everything is correct. That means that you’re taking hundreds, if not billions of dollars and giving hundreds, if not billions of dollars to the Nvidia’s of the world.

So that’s the second space. We have kind of some amount of exposure in that, but not a lot because we tend to think that this will get centralized in the top cloud providers and database providers. Everybody in that space who are existing cloud and database providers will end up providing their own large models, facilities, and things of that nature. However, there will be one or two large companies that will succeed in that space and I’m pretty confident of that, but it’s a very costly endeavor.

The third space is what I call vertical application, which is one layer on top of vertical application building. Here, we have a significant amount of companies that we have picked and invested in based on their expertise in one particular area. For example, we have a company in healthcare. If one goes into an emergency room in a hospital in Stanford, a physician there has to make a decision on a couple of symptoms and takes about five minutes to read from the database the hospital is using. That inherently leads to physicians having trouble to actually diagnose what it is, right? So we provide a tool that helps physicians. It’s created by the architect of Siri, and there’s an MD who was an emergency room physician in its founding team. So there’s a lot of experience within that team to make sure that hospitals and physicians are well taken care of.

While every hospital will say physician burnout is an issue and we should take care of it, when it comes down to it, they won’t pay for anything except for making more money via insurance or making their operations better. So, we think of a physician being able to write those documents, so we expanded the tool to be able to do those services. It’s been deployed in a number of hospitals – just in the first week, it saves $160K of costs. A university hospital deployed it a week ago. So we tend to think of it as sort of an application led and developed by physicians with the expertise in the AI team sitting alongside it and building it.

We have similar things in taxes. We have similar things in a company called Cognitos in Robotic Process Automation (RPA). So we have about eight companies in that light of vertical technology where there’s somebody who has been in that field.

So for example, in this tax company in our portfolio, there’s a former VP of Engineering from Ernst & Young’s who has served in the tax business and franchise business for a long time. So he understands the taxes and how do you really change that in terms of having a vertical application that helps small, medium businesses build their taxes.

We have about eight to ten companies that we have invested in that layer over the last four years. That third layer is what we are focused on because we tend to think that there, it’s a question of expertise of the person that comes in. You can create a much richer flow if you can put a lot of custom data on top of a combination of generic models with private LLMs. So it’s a combination of putting together a feedback loop with essentially your own knowledge basis if you know the space, you can collect the data. So it is a space where we are very keen on investing in. I hope that kind of gives you the market in the three ways that we look at it.

This segment is part 1 in the series : 1Mby1M Virtual Accelerator AI Investor Forum: With Rajeev Madhavan, Clear Ventures
1 2 3 4