1Mby1M Virtual Accelerator AI Investor Forum: Ray Wu, Alumni Ventures AI Fund (Part 4)

Sramana Mitra: Yes, but I was not asking you about AI rollups. I’m asking you about this Y Combinator model of starting a full stack company from scratch.

Ray Wu: I think it depends on how the full stack companies approach the overall go-to-market. We’re starting to see a lot of these kinds of companies. If it’s in the service industry, then I don’t think it requires a lot of capital upfront, because they’re leveraging computing capabilities rather than hiring a lot of people.

If you think about traditional cost structures, it’s mainly the infrastructure cost of computing and the infrastructure cost of the human. Those are the two big fixed costs. Computing costs have already been shrinking. Twenty years ago when in VC, a lot of times you’d fundraise to buy networking, storage, and computing. That’s already come down.

Sramana Mitra: Yeah, but AI has also pushed up computing costs.

Ray Wu: True, but at least it’s a variable cost to a certain degree.

Sramana Mitra: Yes, it’s a variable cost.

Ray Wu: I’m talking about fixed cost—how do you remove the upfront fixed cost? Even human labor, originally very much a fixed cost, can now be modeled more variably. With AI agents, you can test your model from zero to one. Once it works, you go from one to 100. Again, it’s a variable cost—it scales with customer needs. You don’t need a lot of capital upfront.

We’re seeing more innovation and possibilities of people starting with just a few people and building a whole company without needing much funding to test models. Once proven, you can use additional funding for scalability. That’s still where you need funding, because in this attention economy, whether B2C or B2B, there’s limited room for people to remember or access your offering. Reaching them still takes money.

But for early experimentation in a YC-type model, I think the cost structure has gotten better over time.

Sramana Mitra: And I think the whole thesis of what Y Combinator is pushing is based on this agent model that’s developing. People are automating a lot, and if you start from a clean slate, it’s easier to automate many functions and just get things going. That’s very active right now in the startup ecosystem.

Ray Wu: Yes, one hundred percent.

Sramana Mitra: You can get very far with the agent model. What you pointed out is also true—go-to-market is still expensive. You can get something going, get a few customers, but once you enter the realm of programmatic customer acquisition—whether social media advertising, search, or digital advertising—that’s where it is still expensive.

Ray Wu: Yes. I just think that as the VC industry evolves, we’ll see more startups in the early phase. But I still think a lot of things are new—how to scale the company, how to find the right people, how to go to market. That still requires a lot of networks.

That’s the reason we, as Alumni Ventures, still exist—to help with building teams, finding customers, and leveraging networks.

Sramana Mitra: Network helps.

Ray Wu: Network helps, right? That’s what you’re doing—building this great network. In the end, we’re human. It’s the network that matters. But the mundane tasks we do day in, day out, can be automated with agents. I see the scalability of this model. There’s a lot more to discover with the new possibilities of agent infrastructure. We’re very excited about this space.

Sramana Mitra: Very good. Thank you for coming.

Ray Wu: Thank you. Appreciate you having me.

Sramana Mitra: Bye.

This segment is part 4 in the series : 1Mby1M Virtual Accelerator AI Investor Forum: Ray Wu, Alumni Ventures AI Fund
1 2 3 4