Underwriting Platforms, Not Categories

A few months into the new year, I’ve noticed a similar set of questions come up in conversations with colleagues: What are you looking for? Are you investing in consumer or enterprise? Which sectors are you focused on?

The honest answer is all of the above. I’m a generalist, unapologetically so. I loved how an LP described me as a nomadic thematic generalist — equally comfortable backing a live commerce platform, an intelligence software stack for any moving machine, an identity security company, a mathematical reasoning model, or a next-generation semiconductor architecture. What I look for isn’t a specific vertical; it’s founders building companies that can scale far beyond their starting point.

Before I moved into venture, I spent over two decades in capital markets and strategic advisory, leading several major tech IPOs including Facebook, Google, LinkedIn, Alibaba.com, and Yandex, and advising on landmark M&A transactions at Qatalyst. By partnering with these founders, I learned what it takes to achieve massive scale and what platform companies look like: the breadth of their products, the depth of their customer relationships, the compounding nature of their competitive advantages.

The Platform Opportunity

The strongest companies I’ve worked with and invested in share a common trait: they are building toward a platform, creating multiple growth levers across products, customers, geographies, and revenue streams.

Whatnot did not simply launch a live shopping feature — it is building global communities of sellers and buyers across hundreds of categories in multiple countries.

Applied Intuition began as simulation software and has expanded into multiple product lines, building the intelligence layer for everything that moves in multiple industries and leading the way in physical AI.

Axiom is not building another LLM — it is building the mathematical reasoning infrastructure that makes AI verifiable, moving models from plausible generators to trusted systems in industries like semiconductors, financial services, energy, and healthcare. Formal reasoning and verifiability are some of the most important problems in AI.

Unconventional AI is not designing another chip — it is rethinking the computer itself with an AI-first, energy-efficient lens and challenging a 70-year-old computing paradigm that was never built for AI-native workloads.

These companies share a willingness to think much bigger than a single use case or single market. They do not fit neatly into someone else’s TAM framework. They redefine it, and execute to keep expanding the markets in which they can participate.

The Moat Is the Team

One of the defining characteristics of this market is how quickly technical advantages fade. Model capabilities evolve monthly. Tooling improves weekly. Architectures that feel differentiated today can feel commoditized in a matter of quarters. Traditional moats are less durable than they once were.

What holds up is not static IP. It is velocity.

The companies that break out are powered by teams that move faster than the market itself — teams that ship continuously, recognize shifts early, and adapt in real time. They are obsessively customer-focused. They attract exceptional talent.

This is why founder evaluation matters more than ever. I am looking for founders who are intellectually honest about what is working and what is not. Who do not cling to narratives when the data or customers say otherwise. Who build cultures capable of sustained execution at speed — not just bursts of momentum, but compounding velocity.

In this environment, the moat is the team.

Consumer vs. Enterprise Is the Wrong Framing

I am often asked whether I am a consumer investor or an enterprise investor. When I think about the most important companies I have worked with, that distinction rarely holds up for long.

During my time in investment banking, every defining company eventually expanded across customer bases. Google began with search users, but its model evolved to include advertisers, developers, and enterprise cloud customers. Facebook serves consumers, advertisers including small businesses and large enterprises, and developers. LinkedIn’s customers included professionals, recruiters, and enterprises. Amazon started with consumer retail and grew into advertising, logistics, cloud infrastructure, hardware and media/entertainment.

The same dynamic is playing out in AI today. Companies like OpenAI and Anthropic serve individuals, developers, and enterprises. The most ambitious companies are building products that over time will serve multiple sets of customers.

I don’t describe myself as a consumer investor or an enterprise investor because that framing is artificially limiting. What I look for is a product that solves a real problem exceptionally well, and a founder with the ambition and capability to expand from that initial wedge into something much larger. The starting customer is just the entry point. The platform is the destination.

That philosophy shapes which themes I’m prioritizing right now.

What I’m Watching Closely…

Agents Are Forcing an Identity Reckoning

Nowhere is the shift from experimentation to production more visible — or more consequential — than in AI agents. Enterprises are deploying agents that operate inside corporate systems, spin up dynamically, execute multi-step workflows, and interact with sensitive data at scale. Microsoft and IDC estimate there could be 1.3 billion AI agents in operation by 2028. Every one of them needs to be governed and secured.

The problem is that the identity infrastructure most enterprises rely on was built for a human-first world. It was never designed to manage millions of machines and agentic identities operating simultaneously, dynamically, and often without direct human oversight. The attack surface has expanded dramatically, and the tools haven’t kept up. This is what drove our conviction in identity security — and specifically our investment in ConductorOne, which is building a unified identity platform across humans, non-humans, and agents.

Voice AI: A New Primitive

One of the areas I’m most excited about is voice AI — not as a feature, but as a primary mode of interaction with intelligence. It’s also where I’m actively looking at opportunities to make an investment.

For decades, we’ve trained ourselves to type, click, and stare at screens. But speech is how we actually communicate. The gap between how we interact with machines and how we interact with each other has always been a friction point. That gap is closing fast.

What’s different now isn’t just that voice models have improved — it’s that the entire stack has matured simultaneously. Models are more capable, latency has dropped, and infrastructure is reliable enough to deploy in production. That combination is unlocking real value, particularly in customer engagement, support workflows, and anywhere that high-volume human interaction has historically been expensive and inconsistent. Voice isn’t a feature bolted onto an existing product. Done right, it’s a new interface that changes what the product fundamentally is.

I believe a large platform company can be built here — one that doesn’t just improve how we interact with intelligence but redefines the experience entirely. If you’re a founder building in voice AI, I would love to chat.

Giddy Up!

I am so excited to see how the year will evolve. It is, after all, the Year of the Fire Horse — a rare, 60-year cycle event symbolizing intense energy, dramatic transformation, and high-stakes change! 🏇🔥🏇