
Consider Cerebras Systems (CEREBRAS) as a high-conviction play on the shift from general-purpose GPUs to specialized AI hardware, offering speeds up to 18x faster than NVIDIA (NVDA). Monitor the stock following its recent dip toward the $230 range, noting that its unique "dribble lockup" structure may prevent the typical post-IPO sell-off over the next six months. Invest in Planet Labs (PL) to capture the transition from a satellite data provider to an AI-driven "Planetary Intelligence" company, leveraging its dominant 60% revenue share in the defense sector. Look for long-term growth in Space Infrastructure as launch costs drop toward $200/kg, which will eventually make space-based data centers a viable alternative to terrestrial compute. Prepare for a massive IPO cycle peaking in 2026, providing retail investors rare opportunities to buy high-growth tech companies at earlier $1B–$5B valuations.
• Cerebras is a leader in AI silicon, recently going public with a market capitalization discussed around $50–$60 billion. • The company utilizes a "Wafer-Scale" architecture, building a single chip the size of a dinner plate rather than the traditional postage-stamp size. • Key Competitive Advantage: By placing memory directly next to compute on a massive scale, they solve the "data movement" bottleneck. This allows them to run AI models (like those from OpenAI) 15 to 18 times faster than traditional NVIDIA GPUs. • Business Philosophy: CEO Andrew Feldman argues that for AI to be useful, it must be real-time; users will not tolerate "slow search" or waiting for AI responses.
• Architecture over Iteration: Cerebras is positioned as a "domain-specific" alternative to NVIDIA. Investors should note that their success depends on the industry shifting away from general-purpose GPUs toward specialized AI hardware. • Post-IPO Growth: The panel highlighted that the stock opened around $320 and was trading near $230 at the time of the recording. Despite the dip, the sentiment remains bullish on the long-term "secular trend" of rebuilding data infrastructure. • Lockup Innovation: The company is using a "dribble lockup" (releasing shares over six months based on performance hurdles), which may prevent the massive "dump" of shares often seen when traditional 180-day lockups expire.
• Planet Labs operates the world’s largest Earth-imaging fleet (approx. 200 satellites), imaging the entire landmass of Earth every single day. • Revenue Mix: Approximately 60% of their revenue currently comes from military and defense/intelligence sectors, though they also serve agriculture and energy. • The "Large Earth Model": The company is moving from being a "data provider" to an AI company. By feeding their time-series satellite data into AI, they enable "Planetary Intelligence"—the ability to search the physical Earth like Google searches the internet. • Stock Performance: The stock saw a significant 10x move (from $5 to $50) over a 12-month period as the market began to understand the intersection of space and AI.
• Legitimacy through Public Markets: CEO Will Marshall noted that being public is a "legitimizing event" for government and enterprise customers who need to know the company has the capital to survive long-term. • AI Integration: The real value in Planet Labs is not just the photos, but the AI layer on top that can predict floods, fires, or geopolitical threats weeks in advance. • Space-Based Data Centers: A major future catalyst is the potential to move data centers into space. Marshall predicts this will be cheaper than terrestrial centers within 10 years due to 24/7 solar power and falling launch costs (targeting $200–$300/kg).
• The panel predicts 2026 could be an all-time record year for IPOs. • There is a shift in sentiment: companies are moving away from the "stay private forever" mantra and looking to go public earlier (at $1B–$5B valuations) to use the public market as a "sharpening" mechanism. • Insight: Public market investors may soon have access to high-growth tech companies earlier in their lifecycle, rather than only after they have reached trillion-dollar valuations.
• Launch Costs: Costs have dropped 10x in the last decade. This is the fundamental driver allowing for experimentation in space-based compute. • Miniaturization: The "mainframe to desktop" revolution is happening in space. Satellites that used to cost $1B now cost a fraction of that and weigh significantly less while performing better.
• The "Next Moore's Law" is focused on time-bounded compute (speed of delivery) rather than just transistor density. • NVIDIA (NVDA) is the incumbent, but the panel suggests that "share changes" happen when new workloads (like AI) emerge, creating openings for companies like Cerebras or Grok.
• Geopolitical/Regulatory Hurdles: Cerebras faced challenges regarding its UAE investors and CFIUS (Committee on Foreign Investment in the United States) reviews before successfully going public. • Technical Complexity: Moving data centers to space faces significant "last 10%" engineering hurdles, specifically regarding inter-satellite communication clusters. • Market Volatility: While early investors in Planet Labs saw a 10x return, the path was volatile, and the "SPAC" entry method was noted as a different, sometimes more difficult, journey than a traditional IPO.

By @allin
Chamath Palihapitiya, Jason Calacanis, David Sacks & David Friedberg cover all things economic, tech, political, social & poker.