The IPO Comeback: Why Tech Giants Are Finally Going Public | All-In Liquidity IPO Panel
The Signal
Convergence between AI silicon and space-based infrastructure is the central thesis connecting the rise of Cerebras and Planet Labs. While both leaders agree that public status brings liquidity rather than operational breakthroughs, they remain divided on the timeline for space-based compute, with Planet’s Will Marshall betting on rapid orbital adoption and Cerebras’s Andrew Feldman prioritizing architectural divergence over near-term deployment. The dispute over whether space will become a primary compute hub remains largely unsettled, contingent on launch costs reaching a specific per-kilogram threshold.
The Case
- Cerebras, a public AI-chip company, claims their dinner-plate-sized chip architecture is 15 to 18 times faster than incumbent NVIDIA GPUs because it minimizes the bottleneck of moving data between memory and compute.
- Planet Labs, which operates a 200-satellite fleet providing daily global imagery, argues that satellites paired with AI create a 'real-world data layer' for agriculture, energy, and defense, rather than serving a strictly military purpose.
- Will Marshall, CEO of Planet, contends that orbital data centers will become economically viable when launch costs hit $200–$300 per kilogram, a threshold he expects to approach within three years through vehicles like Starship.
- Skepticism persists regarding space compute feasibility; Andrew Feldman notes that building reliable, massive cluster communication networks in orbit remains an unsolved, difficult engineering hurdle.
- Providing a contrarian view on finance, both founders emphasize that the majority of company value accrues years after an IPO, using this to argue against forcing early share distributions that could strip investors of long-term upside.
The 1 Minute Signal Take
This is a high-density, speculative conversation between two founders framing their companies as secular winners in an AI-dominated landscape. The video is worth watching if you want to understand the architectural argument for why Cerebras claims distinct silicon creates an unavoidable advantage over NVIDIA, as the host provides room for that technical nuance that the summary simplifies.
Time saved: