The New PayPal Mafias: Investing in AI’s Next Great Family Trees

Frequently Asked Questions

FAQ

• Where is the transcript for this webinar?

  Speaker 1:
  Welcome everyone. I’m Meera Oak. I’m a partner with Alumni Ventures, and I’m really excited to share what our team has been up to recently. I know we’re at the top, so maybe I’ll give a few seconds for people to sort of get settled a bit. All right, let’s go ahead and get started. We’ve titled this presentation The New PayPal Mafias: Investing in AI’s Next Great Family Trees. And this really was the culmination of a lot of research, conversations with founders, and a lot of pattern matching that have revealed what we feel are the most exciting investment opportunities that we’ve seen this decade.

  Just to call out, I think we all are a bit familiar with the PayPal mafia story, that sort of dynamic that Elon Musk, Peter Thiel, Reid Hoffman, and so many others have gone on to build—an extraordinary collection of world-changing companies—and that all stemmed from PayPal. But we believe that we’re witnessing the formation of something potentially even more powerful.

  So excited to dig in and explore why. But maybe before we get fully started, I might pause and remind everyone we’re speaking today about Alumni Ventures and our views of the associated investing landscape and that this presentation is for information purposes only and is not an offer to buy or sell securities, which are only made pursuant to the formal offering documents for the fund. And if you have any other questions, feel free to visit the URL at the bottom of that page.

  So what are we talking about today? Here’s our roadmap. We’ll start by unpacking the Tech Mafia phenomenon and why these alumni networks consistently produce successful companies and create outsize returns. We’ll look into how AI companies are functioning as an extraordinary founding bootcamp and training that next generation of founders. We’ll explore what makes this AI wave fundamentally different—the structural advantages that previous tech generations simply didn’t have.

  I’ll walk through an OpenAI case study, highlight some companies that I believe are rising stars and becoming talent incubators themselves, and go into the unique advantages that these AI mafia founders have today. And finally, I’ll share our investment approach and explain why we feel there’s a unique window of opportunity right now.

  I might move a bit quickly, but we are recording and we’ll be sending that out. So feel free to go back and rewatch the parts that you may have missed and feel free to book a call with us if you’d like to learn more.

  Now, a bit about me. My background stems from my time at Swarthmore and Dartmouth Tuck School in terms of education, but professionally, I’ve worked across multiple sides of this innovation ecosystem. I held strategic financial and operational roles at Yale. I worked at early-stage venture funds and incubators like Create Venture Studio where I launched businesses and sourced investments across enterprise SaaS, infrastructure, and more.

  And I think having spent the bulk of my time at the earliest stages, I’ve seen how those early innings of company development play out many times, and how those founding teams really form with the right experiences and the right connections. And I’ve seen how they navigate the challenges of building category-defining companies. So that pattern recognition is really crucial for what we’re discussing today.

  But maybe enough about me. I’d love to share a quick overview of Alumni Ventures as well. We launched in 2014 with a pretty straightforward mission: to democratize access to venture capital—opportunities that were previously available only to institutions and large asset holders. So since then, we’ve raised over $1.4 billion in AUM from over 10,000 individual investors and have invested in over 1,500 portfolio companies.

  This level of activity has made us one of the most active venture firms globally and the largest U.S. venture firm focused on individual investors. We have a pretty robust team of about 120 team members, and they’re primarily based in Manchester, but with many innovation hubs across the U.S., including Boston, New York, Chicago, and the SF Bay Area, which is where I’m based.

  And this geographic distribution is one of my favorite features, honestly, because it gives us visibility into emerging trends and founders across the country.

  Now, the trophy case—I won’t dwell here for too long—but I do want to highlight that our approach has earned recognition from really respected organizations in our field. CB Insights has us as a top 20 VC. PitchBook consistently ranks us amongst the most active VC funds, and Fast Company named us as one of the most innovative companies.

  I’m particularly proud of this community that we’ve built. It’s the largest network in entrepreneurship. And this community has created such a powerful engine for identifying promising opportunities early—which is, in part, what we’re talking about today.

  And one question I often get is, who else are we investing alongside? We’ve established relationships with many top venture firms: Andreessen Horowitz, KLA, Sequoia, YC, and NEA. And when it comes specifically to AI investments—which I know is sort of the topic of today—we’ve been particularly active with strategic co-investments.

  Our portfolio companies include Lambda, Together AI, and Unstructured. I could go on and on, but these are core infrastructure players that are enabling the broader AI ecosystem. We’ve partnered with leading AI-focused investors to access these deals early. And that’s often leveraging our network to add value beyond capital.

  The feature that I think a lot of these firms really appreciate is that we’re quite complementary in our nature. Our LP base and the value-add capabilities that we and our network can offer is sort of how we find our way into these really competitive and otherwise inaccessible deals. So really excited about our position in the market and the value that we add to the portfolio.

  Awesome. So let’s get into the heart of today’s discussion. And just to caveat, when I use the term “AI mafias,” again, I’m building on this familiar venture capital concept with some important distinctions about why we think it’s especially unique. So let’s get into it.

  How tech mafias create billion-dollar outcomes.
  The venture world has really long recognized the outsize impact of alumni networks. I mean, look at the name Alumni Ventures to start. But really, what we’ve seen is that there are breakout companies that have stemmed from this PayPal Mafia—as a textbook example: Elon Musk, Peter Thiel, Reid Hoffman, and others who have really gone on to build category-defining companies like Tesla, SpaceX, LinkedIn, Palantir, and countless others.

  And collectively, PayPal—just to take that as an example—their alumni have contributed to over 575 startups and have raised over $200 billion in capital. We’ve seen this playbook unfold among Google alumni and so many others that we’ll sort of get into. But this impact is why this term and these networks have so much notoriety in the field.

  When we take a step back, there are a few common factors that serve as a through line for this phenomenon: hypergrowth, company culture, and trust. And we’ll get into that a little bit more in the next slide.

  So a natural question that we often get is: what is their origin story? How do these mafias form, and what makes these networks so productive? To start, these mafias really emerge from hyper-growth environments that function as what we call entrepreneurial bootcamps. Employees in these organizations learn how to solve new problems, how to build scalable systems, and how to make consequential decisions with limited information. And I think the kicker here is that this can drive hypergrowth, and depending on the company, that can mean millions in revenue, millions in users, or even more. These are important experiences that larger and more traditional organizations rarely provide.

  I think a second attribute is that each successful company also develops a pretty distinctive cultural DNA. In speaking during this process, I had the opportunity to speak with alumni from some of these companies, and PayPal actually has cultivated a bit of a contrarian-thinking and scrappy resourcefulness attitude, while Stripe really emphasized elegant engineering and a really strong developer experience. These traits really become embedded in their employees, and then those employees carry them to new ventures.

  Then finally, and perhaps most importantly, these shared experiences create really dense networks built on trust. I see this in our portfolio companies almost every day. When you’ve weathered serious challenges together, you form bonds that become incredibly valuable. Your former teammate becomes your future co-founder, your previous manager becomes your angel investor, your colleague introduces you to your first enterprise client—and on and on and on. These relationships dramatically accelerate company formation and early growth.

  But yeah, the real question is: what makes AI mafias different and more powerful? I think the next generation of high flyers carries familiar traits but introduces impactful differences.

  First, we’re seeing an unprecedented technical expertise emerge from our founding teams. You have companies like OpenAI, Anduril, and Mistral, which are building teams at the frontier of AI research. These engineers aren’t just skilled in their craft—they grew up with novel architectures and training techniques that are nearly impossible to replicate.

  Second, there’s a mission-driven culture that we’re noticing. Unlike social media or marketplaces, leading AI labs are focusing on very deep questions of intelligence, safety, and societal impact. We think this perspective drives founders to balance that technical ambition from a strong engineering team with really thoughtful consideration of societal effects. That’s becoming an increasingly valuable trait, especially as regulatory focus on AI safety intensifies.

  Finally, we’re entering this perfect time in the tech stack. What I mean by that is the foundation models are built, the infrastructure available to everyone is pretty solid, and there’s a shift happening toward the application layer. It’s sort of mirroring the explosion we saw in the web in 1995 or mobile in 2008. So we feel that AI mafias are really poised to dominate this phase.

  I actually recently spoke with a founder who spent a couple of years at Anthropic working on reinforcement learning from human feedback, and she’s now applying those techniques to manufacturing, which is unlocking entirely new methods for quality control and process optimization. This is pretty fundamental in terms of their breakthrough. It’s not just incremental improvements—it’s a seismic shift for certain industries. So what we’re seeing is this unique blend of unmatched technical depth, mission-driven focus, and the perfect timing, which positions this mafia as pretty different from prior waves of innovation.

  Let me take a double click on OpenAI for a moment because I find their talent diaspora compelling—it offers a window into how these networks operate. Many probably remember that back in late 2022, when ChatGPT launched, it created a cohort of individuals with rare experience deploying AI at massive scale.

  Think about the technical challenges that OpenAI’s employee base solved. They solved for context window optimization, retrieval-augmented generation, hallucination mitigation, prompt engineering, inference optimization—so many technical capabilities that the list goes on and on. This knowledge base is enormously valuable when applied to both vertical and horizontal problems.

  Jumping to an example, we have Thinking Machine Labs, which was founded by former OpenAI CTO Meera Murati. After helping bring both GPT-4 and DALL·E from research to production, she’s now applying those same technical frameworks to new domains. Her insight from OpenAI into model training, deployment, and optimization gives her team a massive—and almost unfair—head start on this front.

  To me, what stands out about this node is how Sam Altman and OpenAI have cultivated, quote, “the family flywheel.” When alumni start companies from OpenAI, he often wires money immediately, then layers in the OpenAI startup fund investments, additional credits for APIs, and access to the organization’s Converge accelerator. So many of these features and value-adds, through this shared infrastructure, create alignment. These startups build on OpenAI’s model, providing valuable feedback and use cases that improve the core technology. This technical knowledge sharing goes both ways. Startups get access to optimization techniques and implementation patterns from OpenAI, while OpenAI gets industry-specific insights from specialized applications. So every alumni that comes from OpenAI really expands the OpenAI footprint. They have more data flows, broader distribution channels, and greater influence on policy and standards. It’s a technical ecosystem that strengthens itself with each new node.

  While model builders are getting the attention—rightfully so—we’re seeing the infrastructure and application layers of the AI stack as equally fertile ground for innovation. Let me highlight some of these rising stars, because we think they’re becoming important talent incubators.

  If we start with infrastructure, we have companies like Lambda (an AV portfolio company), which is building a hardware and software stack for AI training. They are recruiting a team with deep expertise in GPU optimization and machine learning operations—pretty invaluable for the next generation of AI infrastructure.

  In a similar vein, we have companies like Unstructured (also in our portfolio), solving a critical data preparation problem—turning unstructured content like PDFs, HTML, and emails into large language model-ready data. Their team’s expertise in document processing and knowledge extraction represents a technical moat that will spawn numerous vertical-specific applications.

  Just think about any sort of legacy organization that lives on PDFs and paper-and-pen transactions. This is a huge unlock if Unstructured alumni step away and launch companies in those specific industries.

  On the application side, there’s so much to comment on. I won’t go into all of them, but I will call out Cursor, which is building AI-powered coding tools that reached $100 million in ARR in under three years. Their team understands the intersection of development workflows, AI assistance, and a user interface that’s better for almost everyone.

  On the other hand, we have Abridge, which uses speech and generative AI for medical documentation. They’re combining specialized speech recognition with medical knowledge integration to solve a critical healthcare problem.

  What makes these companies particularly interesting as talent incubators is their domain specialization. Unlike previous tech generations that built general-purpose platforms, these teams are developing deep expertise in specific technical niches or industries. When their alumni eventually found startups, they’ll bring both AI engineering knowledge and valuable domain expertise—which is a powerful combination for building targeted solutions.

  Let’s dive into what makes this a bit unfair in terms of the advantage for AI mafia founders. The reality is that this next generation of founders has structural advantages that previous generations simply couldn’t access. I’ll highlight a few that are changing the equation.

  First: foundation models. These are building blocks that have been transformative for startup economics. Think about how Google or Facebook scaled—they had to build their core technologies from scratch. They trained their own recommendation systems. They created their own infrastructure. Today, founders can leverage powerful foundation models as starting points, reducing both capital requirements and time to market.

  I want to share an anecdote because I think it’s staggering. Training GPT-4 cost roughly $100 million and took months on specialized hardware. Today, a startup can fine-tune that same model for their domain in days and with thousands of dollars. This isn’t just iterative improvement—it’s a fundamental shift in what’s possible for smaller teams.

  Second: modern AI developer tooling has offered unprecedented velocity to companies. We have frameworks like LangChain and LlamaIndex, which have made complex AI application development accessible. We have vector databases like Pinecone, which handle data retrieval efficiently. We have evaluation frameworks that help identify and systematically address model limitations. The tooling ecosystem has really accelerated development cycles, allowing teams to build in weeks what previously took months or years.

  Third: I want to touch on what NFX coined recently—“three-person unicorn economics.” Companies like Cursor have reached $100 million in ARR with just 30 to 50 employees, compared to hundreds or thousands at previous unicorns. This capital efficiency creates a virtuous cycle: lower burn rates mean longer runways with less capital, reduced dilution preserves founder ownership, and faster iteration enables rapid product improvement. Smaller leadership teams make decisions more efficiently and quickly. I really appreciate the nuance of what this brings to early-stage teams.

  Fourth: enterprise AI adoption has reached a tipping point. After years of experimentation, large enterprises are now deploying AI solutions that deliver measurable ROI. This dramatically reduces go-to-market friction and accelerates revenue growth. Today’s AI companies are entering environments where buyers already understand the technology and are ready to deploy budget when asked.

  Finally: multidisciplinary founding teams with unique perspectives. Previous tech mafias were predominantly engineering-focused. Today’s AI startups feature teams blending technical AI expertise with domain knowledge. This combination enables them to build systems that aren’t just technically impressive, but also usable, useful, and human-values-aligned. We think that can lead to more enduring companies.

  This chart perfectly illustrates the technological economics revolution I just described. If you look at Cursor’s path to $100 million ARR in just two years—compared to a company like DocuSign, which took nearly a decade to hit that milestone—it’s not just faster growth. It’s a fundamentally different scaling model enabled by AI.

  Again, Cursor did this with 30 to 50 employees. DocuSign needed hundreds. The implications for technological efficiency are profound. AI-powered products can reach more users with less human intervention. Product improvements can be implemented and iterated faster. The result? Exponentially better economics: higher revenue per employee, lower capital requirements, and faster time to value.

  So what does this mean for Alumni Ventures and our playbook? I think given these patterns, we’ve been quite disciplined in approaching this opportunity on Alumni Ventures’ seed team. And to start, we’ve made a pretty concerted effort to identify and build relationships with the strongest technologists, researchers, and operators at these leading AI organizations. We’re monitoring everything from papers, conference presentations, and career movements just to see who might be thinking about making a significant move before it’s known publicly.

  And when someone does spin out, we’ve developed a rapid decision framework. This includes a technical assessment of the founding team’s capabilities, a market opportunity analysis, competitive positioning, valuation, and reference checks throughout our network. We’re being quite methodical, and that allows us to move decisively when these opportunities emerge, even in competitive rounds. And finally, we apply this evaluation criteria across all opportunities. Technical pedigree is always a strong signal, but we rigorously assess all aspects of the business—the market potential, the technical differentiation, and how well the team can execute—because we’ve seen that the best technical talent is incredibly helpful, but it doesn’t always translate to the best businesses. So we need to be holistic when we think about that evaluation.

  Awesome. So why act now? This question is so important. First, we’re entering this critical enterprise AI adoption window between 2025 and 2030. And what I mean by that is: after years of exploratory pilots, large organizations are now deploying AI solutions that deliver measurable ROI. And this really isn’t aspirational—we’re seeing concrete evidence across industries, from financial services firms to healthcare providers to manufacturing companies. People are ready to act.

  From a technical perspective, the barriers to adoption are falling pretty rapidly. Again, models have reached a pretty high performance threshold. We’re seeing tooling that makes integration and scaling way more accessible and achievable. And there’s this technical talent migration occurring, as researchers and engineers leave pretty established AI labs to apply their knowledge to specific problems. The diffusion of expertise accelerates the development of AI solutions across industries, and so it’s creating so many investment opportunities that were not even possible a year ago.

  So finally, AI-native startups can achieve really high revenue figures with headcounts that are measured in the dozens rather than the hundreds. This capital efficiency means that founders can build bigger businesses faster and with less investment than ever before. And we just think that creates a really compelling economic stage for seed-stage investors. The reality here is that the window to participate in this transformation won’t be open indefinitely. So in our minds, now is the time to engage—while these networks are still forming and the most promising opportunities are still emerging.

  Awesome. So if this philosophy resonates with you, I’d really love to continue the conversation. Our Alumni Ventures Seed Fund is specifically designed to capitalize on these emerging networks with a disciplined investment approach. And we have a fantastic team—I know I’m a little biased, but I think we’re incredible—and we’re geographically positioned to build relationships with the next generation of founders. I’m based in the SF Bay Area, and my colleagues are based in New York and Boston. So we are sort of positioned for where this innovation is actually unfolding. If you’re interested in scheduling a call or would like to access our fund materials, feel free to utilize the QR codes here.

  Awesome. And before we wrap, we did get a few questions submitted ahead of time as well as some that came through live. So let me just pull those up and we can get to them.

  Alright, so one question came through that was asking: how are these AI mafia startups navigating the increasingly complex regulatory landscape, especially regarding AI safety, data privacy, and competition concerns that weren’t as prevalent for previous tech mafias?

  Yeah, this is such a great question and something we discuss pretty regularly. AI mafia founders are taking a pretty different approach from previous tech waves because they’re prioritizing safety and responsible deployment from the start. By being proactive, features like audit trails and governance frameworks are being embedded from the get-go, and it really offers a competitive edge. In heavily regulated industries like healthcare and finance, this commitment to trust and transparency has only accelerated adoption.

  So with this mindset, we’re seeing that advanced techniques like federated learning and synthetic data generation are minimizing risk and positioning companies for long-term success as regulations evolve. That’s a great question, though.

  And then we have another one: are these AI mafias primarily a Silicon Valley phenomenon, or are you seeing similar talent networks forming in other global tech hubs, and how does that affect investment strategy?

  Yeah, this is a great question as well. I’m based in Silicon Valley, and so I’ve certainly found that the San Francisco Bay Area has the highest concentration of these networks. But that said, AI mafias are popping up all over the world, and it’s definitely shaping how we think about investing.

  If you just look to Europe—London, with its DeepMind alumni, is becoming a hub for scientific AI. Paris is making waves in foundational model development thanks to Mistral AI and their strong academic roots. We’re seeing the same thing play out in Toronto with the Vector Institute, which is fueling innovation in healthcare AI. And Israel still remains a leader in AI security and infrastructure.

  Actually, just the other week I spoke to a founder based in Stockholm who recently spun out of Vev, which is this up-and-coming mafia in the web creation space that has grown tremendously over the past year. So our strategy is evolving to match this global shift. We’re building relationships in these key hubs, we’re understanding the local talent dynamics, and we’re adjusting our evaluation frameworks to fit each region’s strengths.

  I think that might be all the time that we have, and I really want to thank you all for joining us today. I hope I’ve conveyed why we’re so excited about AI mafias and the transformative companies that they’re creating. If you’re interested in this topic, please keep an eye out for more content here. And thank you again for joining us.