The IPO Comeback: Why Tech Giants Are Finally Going Public | All-In Liquidity IPO Panel
Episode Details
The 'All-In Liquidity IPO Panel' features a discussion on the rebounding IPO Market, moderated by Jason Calacanis. The panel includes Andrew Feldman, founder and CEO of Cerebras; Will Marshall, co-founder and CEO of Planet Labs; and Brad Gerstner, founder and CEO of Altimeter. The conversation starts with an anecdote about Davos, where Jason Calacanis covered an interview with Donald J. Trump after receiving a call from David Sacks. The core discussion centers on the transitions of Cerebras and Planet Labs from the Private Markets to the Public Markets. Planet Labs went public via SPACs, significantly rewarding its early Venture Capital backers like Founders Fund, founded by Peter Thiel, DST, founded by Yuri Milner, Capricorn, and Google. To manage newly public transitions safely, innovations like the Dribble lockup are being implemented to distribute shares thoughtfully, a concept Altimeter discusses and supports. Technologically, the panel explores breakthroughs in AI Chips and Artificial Intelligence (AI). Andrew Feldman notes that Cerebras intentionally avoided building a standard GPU like Nvidia or legacy processors like those from Intel, AMD, and Arm. He REFERENCES Moore's law and notes that true innovation requires Domain Specific Architectures that allow partners like OpenAI to run inference significantly faster, much like how networking transformed with Cisco, Juniper, and Arista. Similarly, Groq is taking an alternative silicon path. In the space sector, Will Marshall details how Planet Labs is revolutionizing Earth imaging to provide planetary intelligence to clients like NASA. He foresees the migration of compute power to data centers in space, leveraging cheap Solar Energy and cooling. This shift depends heavily on lowering the Cost per kilogram to orbit, an effort spearheaded by Elon Musk and SpaceX through systems like Starship. The ability to transport massive data from space is also bolstered by networks like Starlink and OneWeb. In the near future, Google plans to launch its TPU hardware into space for early tests with Planet Labs. Ultimately, the panel emphasizes that combining real-world data from space with Large Language Models (LLMs) developed by companies like anthropic and OpenAI will create massive new ecosystems in both the hardware infrastructure of terrestrial Data Centers and global networking.
The episode highlights a shift in the IPO market, where high-growth technology companies are increasingly choosing to go public earlier to access capital and legitimacy. Key investment themes include the evolution of AI silicon through domain-specific architectures and the emerging potential for space-based data infrastructure as launch costs decline.
Portfolio lens: This set of ideas represents a high-conviction bet on the next generation of physical infrastructure, focusing on the hardware and spatial layers that will support the next wave of AI and global data intelligence.
Generated with gemini-3.1-flash-lite on 7/19/2026, 5:38:44 AM. For research only. Not financial advice.Domain-Specific AI Silicon Infrastructure
The next phase of AI hardware growth will be driven by domain-specific architectures that prioritize memory-to-compute efficiency over general-purpose GPU designs.
The episode argues that general-purpose GPUs are hitting diminishing returns and that specialized architectures, like those from Cerebras, are necessary to solve the data-movement bottleneck in AI inference.
- Cerebras claims their architecture is 15-18 times faster than standard GPUs for specific AI workloads.
- Historical precedents like the shift from mainframes to desktops and the rise of networking hardware companies like Cisco and Arista suggest that new workloads create opportunities for non-incumbent silicon providers.
- Increased enterprise demand for real-time AI inference.
- Adoption of non-GPU silicon by major AI labs like OpenAI.
- High capital expenditure requirements for chip manufacturing.
- Incumbent dominance of Nvidia and their massive software ecosystem moat.
- Analyze unit economics and power efficiency metrics of domain-specific chips versus standard GPUs.
- Monitor enterprise adoption rates for non-Nvidia hardware in inference-heavy applications.
Space-Based Compute and Earth Intelligence
As launch costs continue to decline, the infrastructure for planetary intelligence—combining real-time satellite imagery with space-based data centers—will become a viable, high-growth sector.
The combination of lowering launch costs via SpaceX and the ability to leverage solar energy in space for cooling and power makes non-terrestrial data centers a potential long-term infrastructure play.
- Launch costs have decreased by 4-5x over the last decade.
- Planet Labs is already testing the deployment of Google TPUs and Nvidia GPUs in space.
- Space-based solar power provides 24/7 energy access without the need for batteries or terrestrial grid reliance.
- Successful deployment and scaling of SpaceX Starship to further reduce cost per kilogram to orbit.
- Integration of Large Language Models (LLMs) with real-time Earth observation data.
- Technical challenges in building and maintaining compute clusters in space.
- Latency and bandwidth constraints for data transmission between space and Earth.
- Long-term regulatory and orbital debris management issues.
- Track the cost-per-kilogram-to-orbit metrics for SpaceX Starship.
- Evaluate the progress of early-stage space-based compute tests by companies like Planet Labs.
Watchlist
- Cerebras
- Planet Labs
- SpaceX launch cost metrics
- Nvidia
- Google TPU deployment status
Open Questions
- At what specific cost-per-kilogram threshold do space-based data centers become economically superior to terrestrial ones?
- Can specialized silicon providers build a software moat comparable to Nvidia's CUDA?
- How will the 'dribble lockup' mechanism affect the volatility and liquidity of newly public tech companies?
Key Topics & People
The financial market for Initial Public Offerings, experiencing major events with AI companies.
The concept of hosting AI data centers in orbit to circumvent terrestrial energy and zoning constraints.
.jpg?width=300){kind=avatar}
Large facilities housing servers and GPUs for AI compute, which are increasingly hard to power and zone.
Investor and podcast host offering insights on geopolitical conflicts, US policy, and politics.
.jpg?width=300){kind=avatar}
Investor and podcast host moderating discussions on startups and tech markets.
Investment funding provided to startups and early-stage companies.
A venture capital firm known for highly concentrated, massive returns.
Renewable power generation that consumes significant amounts of copper and faces a looming bottleneck in silver supplies.
Technological field experiencing a super cycle in infrastructure, models, and secondary investments.
Investor and host on the podcast who presents data on secondary markets and liquidity.
Co-founder and CEO of Planet Labs, a company indexing the earth via satellite imaging.
An earth imaging company that operates a massive fleet of satellites to provide global daily data.
Custom-designed silicon optimized for specific workloads rather than general-purpose computing.
An innovative IPO mechanism allowing early investors to slowly sell shares according to performance hurdles.
The process of capturing high-resolution daily imagery of the planet via satellite networks.
Observation regarding the density and capability of silicon processors over time.
Venture capitalist whose firm, Founders Fund, invested in Planet Labs.
Prominent technology investor and backer of Planet Labs through DST.
Former US President mentioned in relation to an interview at Davos.
Financial markets for publicly traded equities, where liquidity and large-scale growth can occur post-IPO.
Financial markets for unlisted companies, where historically most venture value was captured.
The key economic metric dictating the feasibility of space-based businesses and data centers.
Advanced AI models trained on massive text data, poised to integrate real-world physical data.
.svg?width=300){kind=logo}
.jpg?width=300){kind=avatar}
Founder and CEO of Cerebras Systems, an AI silicon company.