Paul Murphy

Mega seed or starve: The Bifurcation

The core message of the provided transcript is that early-stage AI funding is splitting into two very different paths: a small set of companies raising extremely large seed rounds at rapidly rising valuations, and many others facing a much harder fundraising market. The moderator frames the panel around new data attributed to Carter showing that the top 5% of seed round valuations rose from $70 million to more than $200 million between Q1 2025 and Q1 2026. Filip from Cherry Ventures describes this as a bifurcation: hot AI companies are attracting larger rounds, while companies outside that category struggle. He argues that round size should not be treated as a vanity metric, but as a function of capital necessity, especially where compute, hardware, deep tech, or long R&D cycles are involved. Rob from Visionaries challenges the idea that AI companies inherently need more capital. He argues that AI also makes company-building more efficient, especially for application-layer startups. He cites portfolio examples including Langnung, which he says reached $70 million ARR after raising only $3.5 million, and Lovable, which he says reached $50 million ARR with only 25 people. His practical advice is for founders to raise only what they need, choose investors for their ability to add value, and preserve ownership where possible. He warns that large early rounds can create narrow outcomes for founders if liquidation preferences and high valuations leave little value at exit, even in what would normally be a strong billion-dollar sale. Paul Murphy, recently departed from Lightspeed, places mega-rounds in the context of an early AI infrastructure buildout. He says large rounds can make sense when two conditions are present: the company can productively use the capital, for example on GPUs, and the founding team represents scarce talent. He uses Mistral as an example, saying the investment case rested on the belief that the three founders could do something few others could do at that moment. However, he warns that the market may now be conflating genuinely rare “n of one” teams with a broader pool of talented people from strong institutions, which could lead to a shakeout in the coming years. The discussion then returns to founder decision-making. Filip emphasizes that raising too much too early creates dilution, sets a demanding price for the next round, and can crowd out future investors who cannot lead at inflated valuations. He distinguishes between application AI companies, which may reach significant revenue with lean teams, and hard deep tech companies such as Proxima Fusion or exploration companies, where hardware, long cycles, and heavy R&D justify larger capital needs. The transcript provided ends during a follow-up question at 13:48, so later session content is not available for analysis.