TechCrunch’s StrictlyVC soirée in Los Angeles last week brought together two AI investors who are refreshingly honest about how little anyone knows right now. Carter Reum, co-founder of M13 (that's $2.5 billion in assets under management and 17 unicorn investments, if you're counting), and Chang Xu, a partner at Basis Set Ventures (nearly $1 billion in AUM, fourth fund, all AI all the time), sat down in a sun-drenched El Segundo room to explain the unexplainable.
Xu kicked things off by declaring the market both a bubble and not a bubble - a paradox that would make Schrödinger's cat proud. ChatGPT went from zero to $40 billion in revenue in six months, which is apparently what happens when you stop worrying and learn to love the hype. Meanwhile, Basis Set portfolio company OpenArt rocketed from $1 million to $70 million ARR in two years with just 20 people and a cash-flow-positive attitude. The bar for 'good growth' has been thrown out the window, and valuations only look insane if you don't assume everything will compound forever. But if you do assume that for every deal, your portfolio will explode. So: same as always, but faster.
Reum brought historical perspective, noting that panic over new technology is as old as the car (1920s, people lost jobs, life went on). What's different this time? It's not just innovators versus innovators - it's innovators versus the ten largest, most well-funded tech companies on Earth. For the first time, incumbents might actually have the edge: tech, capital, data, talent. So startups can rise faster than ever, but they can also fall faster. Reum finds this harder to invest in, but hey, if you get it right, you look like a genius. No pressure.
On pricing deals when revenue appears overnight but sustainability is a ghost: Reum does cocktail napkin math. He recently passed on an AI software-for-brands startup because the numbers didn't check out. Xu focuses on technical defensibility, which changes every quarter, month, or week. The hot new thing? 'GitHub for agents.' Last year that didn't exist; this year there are at least ten teams chasing it. Above the AI layer, it's all about long-term differentiation. Below the AI layer, infrastructure is being rebuilt because databases and version control were designed for humans, not agents. Agents, it turns out, are picky.
How to avoid getting steamrolled by OpenAI, Anthropic, or Google? Reum loves regulated industries - friction as a moat. His firm had a near-billion-dollar exit in AI for 911 call centers. Hyperscalers will get there eventually, but not for a few billion dollars. Healthcare is also on the list: they'll come, but regulation slows them down. Xu distinguishes between 'velocity markets' (fast followers win) and 'depth markets' (hard things stay hard). Example: a portfolio company using transgenic chickens to manufacture complex proteins. Chickens still take time to hatch, so that's a depth market. For now.
Are there genuinely novel ideas, or just remixes? Both, says Xu. Consensus categories like agents for finance and healthcare have strong founders who will likely win. But the most interesting ideas are the ones you can't imagine as a business - like OpenArt, which started as a prompt-discovery page after Dall-E and Stable Diffusion launched. How is that a business? Nobody knew. Two years later: $70 million ARR. The lesson: bad ideas become good again. Four years ago, selling to Hollywood was a bad idea. Then generative AI happened. Now Cursor, dismissed as an 'AI wrapper,' exits at $60 billion. And researchers, once paid barely above the poverty line, are Twitter celebrities.
Reum thinks we're in the early innings. The first wave is obvious and crowded; the second and third ripples (like a heavy rock skipping across water) are where the magic happens. Those are harder bets but come with saner valuations and better ROIs. The SpaceX IPO, meanwhile, is about to shower LA employees with cash - more widely distributed than any VC-led liquidity event. Reum advises anyone with a house, boat, or plane to take advantage. More importantly, major liquidity events spawn second waves. The previous LA cycle gave us Riot Games, Tinder, Snap. This one could be bigger.
Three years ago San Francisco was declared dead. It's not. But LA has something SF lacks: taste. The next frontier in AI isn't more compute - it's making things that resonate emotionally, culturally. SF has the technical talent; LA has taste in spades. And as AI automates more technical work, taste becomes the differentiator. So maybe the chickens - transgenic or otherwise - will come home to roost in Los Angeles.