The IPO Comeback: Why Tech Giants Are Finally Going Public | All-In Liquidity IPO Panel

The panel opens with Brad Gerstner introducing Andrew Feldman (Cerebras) and Will Marshall (Planet Labs) as founders of two consequential newly public companies representing major trends in AI silicon and space data. Jason Calacanis briefly recounts his Davos experience moderating an interview with President Trump before the substantive discussion begins.

Andrew Feldman, whose Cerebras IPO priced at $185 and opened at $320 (reaching a ~$50-60B market cap), offers a grounded perspective on going public: while employees celebrate and families feel proud, nothing fundamental about the business changes overnight. Vendors, engineering progress, and core operations remain exactly the same. He notes the IPO process itself is filled with low-value-added administrative work but acknowledges it provides capital and employee morale validation. Gerstner adds context that Cerebras faced particular challenges going public due to UAE investor involvement and CFIUS scrutiny under the Biden administration, making the eventual smooth IPO feel like '9.5 years of hard followed by 12 months of easy.'

Will Marshall contrasts Planet Labs' experience — going public via SPAC in 2021 at roughly $2 billion with minimal market attention, then watching the stock 10x from $5 to $50 over the past year. He argues going public provides legitimacy and permanence that matters enormously to enterprise customers like governments and large agricultural companies who need assurance the company will remain operational. Most early investors including Google held their shares through the entire appreciation period, capturing the majority of value in the public markets.

On space-based data centers, Marshall explains Planet's research with Google showing that when launch costs fall to $200-300/kg (currently ~$1,000/kg, down 10x over a decade), it becomes cheaper to put compute in space than on the ground. The key advantage is solar power: satellites in sun-synchronous dawn-dusk orbits receive 24/7 sunlight, generating 5x more energy per solar panel without needing batteries or backup power. Planet is already launching Nvidia GPUs and Google TPUs into space as early tests. Marshall believes most compute will be in space within 10 years. Feldman respectfully pushes back, noting that building distributed compute clusters in space is extremely hard when we haven't fully solved it on the ground, comparing the challenge to how autonomous vehicles took far longer than expected.

On AI silicon, Feldman explains that AI fundamentally expanded what computers can do — enabling image understanding, language generation, and comprehension that was previously impossible. This opened massive new compute markets. Cerebras' architectural bet was that if you want to be 20x better than Nvidia, you cannot build a GPU — you need a fundamentally different design. Their solution was a chip the size of a dinner plate (versus postage stamp-sized GPUs) with memory placed directly adjacent to compute, enabling much faster memory access. This makes Cerebras 15-18x faster than GPUs for inference, which Feldman argues is critical because users will not wait for slow AI responses, just as they abandoned dial-up internet and slow search.

The panel closes with a broader discussion about IPO timing and public market value creation. Gerstner and both founders argue that historically more value is created after IPO than before, and that the decade-long 'stay private forever' philosophy championed by firms like Andreessen Horowitz is swinging back. Planet Labs is cited as proof that 10x returns can happen in public markets. The speakers suggest the era of mega-private companies like Anthropic, OpenAI, and SpaceX accumulating all value pre-IPO may not be the new normal, and that earlier public listings — even at lower valuations — sharpen focus and democratize participation. Marshall closes with a vision of 'planetary intelligence' — AI models fed real-world earth observation data rather than just internet text — as the defining technological opportunity of the next decade.