AI early-stage investing has never been harder to price, and two of the more candid voices working in the space said as much at TechCrunch’s StrictlyVC evening in El Segundo last week. Carter Reum, co-founder of M13, and Chang Xu, a partner at Basis Set Ventures, spent an hour on stage dissecting how to find defensible companies when the competitive landscape can shift in a week.
Reum’s firm, M13, manages $2.5 billion in assets and counts seed or Series A positions in 17 unicorns. A Form D filing for M13 Ventures IV disclosed a $400 million offering in 2023, with Carter Reum listed among the related persons. Xu’s firm launched in 2017 as one of the first early-stage funds to focus exclusively on AI. In January 2026, Basis Set closed its fourth fund at $250 million, up from a $185 million third fund, bringing total assets under management to $850 million, according to the firm’s own announcement; the Wall Street Journal also covered the raise. Known limited partners include Foundry and Melinda French Gates’ Pivotal Ventures.
Pricing Deals When Growth Curves Break Every Model
Xu invoked Basis Set portfolio company OpenArt to illustrate how drastically the benchmark for strong growth has shifted. OpenArt scaled from $1 million to $10 million in annual recurring revenue in its first year, then from $10 million to $70 million in year two, remaining cash-flow positive for most of that period with just 20 people. The company has since raised a $30 million Series A led by Canaan, and at the time of that announcement reported more than 8 million monthly active users and $70 million in ARR, according to Basis Set’s own account of the round.
Xu also cited ChatGPT’s revenue trajectory as evidence that terminal-value assumptions have been rewritten. She put the figure at going from $1 billion to $40 billion in revenue in six months. OpenAI’s CFO Sarah Friar has separately put the company’s full-year 2025 annual recurring revenue above $20 billion, according to Sherwood News; the gap likely reflects different time periods or revenue definitions rather than a single reconcilable number. Either way, the directional point stands: the scale of growth now observable makes even stretched valuations defensible in isolation.
‘When you have this possibility of compounding accelerant growth, the valuations don’t seem so crazy because you price that into the terminal value,’ Xu said. ‘On the other hand, if you price every single deal to that math, there’s no way that will work out well for a portfolio.’
Reum’s approach is blunter. ‘We always do the cocktail napkin math,’ he said, describing a recent pass on an AI software business for brands. The question he asked: how big were the category winners last cycle, are there going to be more brands in the world, and are they willing to pay double or triple for software this time around? The investment didn’t clear the bar.
The AI Early-Stage Investing Case for Depth Over Velocity
Both investors have built frameworks for distinguishing companies that can survive hyperscaler encroachment from those that cannot. Xu draws a line between velocity markets, where fast followers win on execution speed, and depth markets, where hard problems remain hard regardless of who has the most compute. A Basis Set portfolio company using transgenic chickens to manufacture complex proteins sits firmly in the latter: biological timelines are not compressible by capital.
Her other framework is structural. ‘Below the AI’ covers infrastructure being redesigned from scratch because it was built for humans, not agents: databases, version control, deployment tooling. ‘Above the AI’ means application-layer businesses where the moat is long-term differentiation rather than a temporary technical lead. The boundary between the two moves every quarter.
Reum’s equivalent is what he calls friction as a moat. M13 had an exit of just under $1 billion in a company that disrupted 911 call centres with AI, a market regulated tightly enough that the hyperscalers have little incentive to enter at that outcome size. Healthcare sits in a similar category: the incumbents will arrive eventually, but compliance timelines buy runway.
‘You used to see them coming in the rearview mirror,’ Reum said. ‘I tell every founder: you need a microscope in one eye and a telescope in the other.’
On the longer arc, Reum is straightforwardly bullish about Los Angeles, framing the anticipated SpaceX IPO as a liquidity event with unusually broad distribution among employees rather than a concentrated windfall for institutional investors. The previous LA cycle produced Riot Games, Tinder, and Snap. He believes what follows the current technical wave, built on brand fluency, content, and cultural understanding, will be centred here. Xu put the same argument more precisely: the next frontier in AI is not more compute, it is taste, and that is a resource San Francisco does not monopolise.
For anyone watching the WSJ’s coverage of the AI venture landscape, the more immediate test is whether the second and third ripple bets Reum describes can be identified before consensus catches up. His view: fewer people are thinking about them, valuations are more reasonable, and the returns, historically, are better. The catch is that those are also the bets that look the most wrong before they look right.
