Placing the Right Bets: Sancus' Framework for Venture Investing

Great investors don't chase trends—they find signal amid the noise. I had the privilege of sitting down with Lake Dai, Founder and Managing Partner of Sancus Ventures, to understand her framework for placing the right bets in AI investing. Lake’s reputation precedes her, and Sancus Ventures was recently selected by the AngelList Systematic Fund-of-Funds for investment.

Lake's unique perspective on early-stage investing is shaped by her expansive background within the startup ecosystem, as a researcher, founder, and venture capital operator. With pivotal early roles at Alibaba and Yahoo! China, more than two decades of technology leadership experience in machine learning innovation in Silicon Valley, and as founder and Managing Partner of Sancus Ventures, Lake brings a wealth of knowledge to her investment approach.

Responses have been edited for clarity and length. Full conversation can be found here.

What first intrigued you about venture capital and led you to start Sancus Ventures?

Lake: Although my journey has involved large corporations like Alibaba and Yahoo! China, in my heart, I've always been drawn to startups. As I continued working in the startup world, I realized we needed VCs who really understand deep tech and have hands-on experience building large-scale infrastructure from day one.

I started Sancus Ventures because I wanted to build a firm that I would want to raise money from when I was a founder—one that provides critical value for startup founders at the stage of investment.

There's always been high demand to join your funds, and your LP base includes retail investors, institutional investors, and is international in scope. I understand you're somewhat selective in who you allow to invest in Sancus. Can you tell us about how you approach fundraising?

Lake: Fundraising can be incredibly difficult. I compare it to being a founder—you can have an excellent product and confidence in your company, but you still need to be very strategic about which investors you want to work with. The same applies to LPs.

For us, we look for what we call "long, kind, and strategic" investors. First off, we look for those with a long-term perspective and patience. We're very honest with our investors that, compared to hedge funds or other assets, VC is generally a very illiquid asset class, so patience is important.

We also look for LPs who are a cultural fit with mutual respect. While we of course want to bring financial returns, we also want to see strategic synergies. As a result, we have a number of multi-billion tech founders and prolific family offices as our LPs. Our strong network of VPs, CTOs, CIOs, and engineering leaders are our portfolio companies' customers, as well as our due diligence partners and LPs.

You have some very impressive startups in your list of portfolio companies. What makes Sancus different in terms of accessing high-quality founders and deal flow?

Lake: This is like asking Coca-Cola for their top secret recipe! Joking aside, we hold a very high standard for both deals and founders. We love founders with proven track records. In fact, our Fund I portfolio companies include six unicorn founders, including the founders of Databricks, Twitch, Freenome, and more!

There are few things that I think have solidified us as a firm that founders want to work with.

1. We deeply understand what they're building. Typically, when founders talk to VC firms, they might talk to a scout, then a partner, then a General Partner who brings in a specialist, and by the fourth meeting, the firm finally understands their technology, differentiation, mission, and vision. With us, because we are deep in the industry and network, we immediately get it. The conversation then becomes what we call "brainstorming the product roadmap" or "co-engineering the future."

2. Everyone on our investment team has been a founder and has built products in the past, so there's often an immediate connection because we truly empathize with and understand what they are going through.

3. We also bring immediate impact through critical customer introductions. Potential customers participate in our due diligence process, which not only helps us have more sophisticated due diligence but also directly connects founders with potential customers.

You actively teach at Carnegie Mellon University as an adjunct professor. How does your academic work relate to what you do at Sancus Ventures?

Lake: I love teaching. I’ve been an AI professor at Carnegie Mellon for nine years now. My role allows me to work closely with Stanford, MIT, Berkeley, and other universities that are advanced in AI research, which helps me see trends from academia's point of view.

In my courses, I focus on real-world applications, like building and tuning machine learning models for manufacturing, retail, healthcare, finance, and more. I co-teach with 10-20 industry leaders each year, including CTOs and CIOs from Fortune 500 companies, tech leaders from Google and NVIDIA, and startups working on real problems. This community of CTOs, CIOs, and startup founders helps me keep my finger on the pulse of the industry and provides a network that benefits our portfolio companies.

Given your expertise, what are some key factors that have driven to the seemingly rapid rise in AI?

Lake: I actually saw a pivoting point for AI development back in 2016. For those of us who have been operators in the machine learning space, we know that machine learning and large language models existed way before ChatGPT. What ChatGPT has been especially successful at is helping the media and everyday consumers understand how important and powerful AI can be.

I see three major pivoting moments or resources.

First is the availability of data—the explosion of digital data from smartphones, smart devices, and more massive amounts of text, video, and images for training.

Second, 2016, specifically, was a pivoting point because we saw GPUs start to really accelerate. Previously, most people understood NVIDIA as a gaming chip company or for crypto usage, but that year we started seeing AI startups creating use cases using distributed computing and parallel computing with GPUs.

The third element is algorithms. We've seen breakthroughs in reinforced learning and new methodologies to train models more efficiently. Over the last eight years, we've seen many algorithm developments on the training side, and now we're seeing more on the inference side.

What's next for Sancus Ventures as you move from Fund I to Fund II?

Lake: We'll continue building on our reputation among founders as we increase Fund II's size. The important thing isn't an arbitrary fundraising goal—it's ensuring the fund size supports our strategy and enables even better returns for investors. We're refining our approach to check sizes, ownership targets, and valuation guidelines.

Our strategy—putting in first checks and then increasing ownership by closely monitoring companies—has been described as something people haven't seen in the market before. With Fund II, we're focusing on improving investment returns through increased ownership and disciplined valuation.

Looking ahead, we expect to see continued acceleration in AI adoption, which is why we're particularly interested in bringing in strategic LPs who can serve as use case adopters and go-to-market partners for our portfolio companies.

Conclusion

For additional insights or to share the conversation with others, you can find the full conversation with Lake, including a live Q&A from the audience, below.

We’re proud to partner with Sancus Ventures as they continue to grow with each new fund. To learn more about Sancus Ventures, follow them on LinkedIn or reach out directly to Sancus at hello@sancus.vc.

Curious to learn more about how they and 25K+ funds and syndicates are leveraging AngelList? See all of our offerings for funds here.