TLDR
A panel discussion with CEOs of two newly public tech companies — Cerebras Systems (AI silicon) and Planet Labs (Earth observation satellites) — alongside investor Brad Gerstner. They discuss the realities of going public, the convergence of AI and space technology, the future of space-based data centers, and why most value creation happens after IPO rather than before. Key themes include Cerebras' wafer-scale chip architecture delivering 15-18x speed over GPUs, Planet's vision for 'planetary intelligence' combining satellite data with AI, and the economic tipping point for orbital data centers as launch costs plummet. Both CEOs emphasize that public markets sharpen focus and that investors should hold shares longer post-IPO. Will Marshall predicts most compute will move to space within 10 years, creating a multi-trillion dollar market. Andrew Feldman argues AI speed is existential: 'How big is the market for slow search? Zero.'
For you
PRO members can sign in and request more information — or their own version of this digest, tuned to what they care about.
Sign in to unlockSurprising Insights
Planet Labs derives ~60% of revenue from military/security customers — far higher than its public 'Earth imaging for agriculture/climate' branding suggests.
A wafer-scale chip the size of a dinner plate can be manufactured with acceptable yields; Cerebras solved the defect-tolerance problem that killed prior attempts (e.g., Trilogy Systems).
Space data centers don't need nuclear, gas, or batteries — just solar panels in dawn-dusk orbit providing continuous power, making orbital compute infrastructure radically simpler than terrestrial.
The 'last 10% of the problem takes 80% of the time' dynamic (like self-driving) may apply to space-based compute clusters; inter-satellite optical links remain a major unsolved challenge.
Brad Gerstner notes the pendulum is swinging back from 'stay private forever' (Andreessen era) to earlier IPOs at $1-5B valuations, letting public markets capture the 10-100x growth previously reserved for private investors.
Key takeaways
Going public is operationally anticlimactic — core business doesn't change, but external validation helps with customers, recruiting, and credibility; most value creation occurs post-IPO.
Space-based data centers become economically viable when launch costs reach $200-300/kg (currently ~$1,000/kg); Starship could hit this in 2-3 years. Solar panels in sun-synchronous orbit yield 5x energy with no batteries needed.
Cerebras' wafer-scale chip (dinner-plate size) places memory directly next to compute, solving the data-movement bottleneck and delivering 15-18x faster inference than GPUs — speed is existential for AI adoption.
AI models are 'blind' to the physical world; fusing real-time planetary sensor data (Earth observation) with LLMs creates 'large Earth models' / 'planetary intelligence' unlocking massive new application categories.
Investors systematically underestimate post-IPO upside: studies show more absolute dollars are made after IPO. Lockup innovations (dribble releases) can align incentives. Planet's 10x public-market run (from $5 to $50) exemplifies this.
Notable quotes
“We are participating in something extraordinary. On everything we do, we are the fastest bar none.”
▸ Notable quotes · 0:11“Space and AI are really a match made in heaven. They're getting married, in fact. Just like Google figured out how to index the internet and make it searchable, we are indexing the earth and making it searchable.”
▸ Notable quotes · 0:20“I believe that the wave is the biggest wave in the history of technology. It will be incredibly beneficial for America.”
▸ Notable quotes · 0:40“The hard part here, the hard part, is moving data from memory to compute. This is the fundamental problem in AI.”
▸ Notable quotes · 23:19“How big is the market for slow search today? Right? It's zero. How big is the market for dial-up? It's zero. You will not wait for AI. We have to deliver it to you in real time.”
▸ Notable quotes · 24:21“I call them blind. You know, they don't know about that farm field, that flood, the security situation around the corner. If you give them real world data, then they can answer real world problems.”
▸ Notable quotes · 31:36“Instead of having large language models, large Earth models. Or instead of AI, planetary intelligence, where you have planetary sensing systems in space, planetary compute systems in space.”
▸ Notable quotes · 31:52Chapters
0:00 — Introductions and Davos anecdotes: Jason Calacanis opens with banter about Davos, a Trump interview, and introduces the panel: Brad Gerstner (Altimeter), Will Marshall (Planet Labs), Andrew Feldman (Cerebras).
2:00 — The reality of going public: Feldman (3 weeks post-IPO) and Marshall (~4 years post-SPAC) describe the IPO process as grueling but operationally inconsequential — core business unchanged, employees celebrate then return to work. Gerstner adds context on Cerebras' difficult path (CFIUS, UAE investor) and explosive debut (priced $185, opened $320, now ~$230, $50-60B market cap).
13:20 — Space-based data centers: the next trillion-dollar market: Marshall details Planet's study with Google: at $200-300/kg launch cost, orbital data centers become cheaper than terrestrial. Starship could hit this in 2-3 years. Sun-synchronous orbit gives 5x solar energy with no storage needed. Planet already launching NVIDIA GPUs and Google TPUs as tests. Marshall predicts most compute moves to space within 10 years.
19:20 — Cerebras and the rebirth of AI silicon: Feldman explains AI unlocked compute for images/language (previously 'bad at'). New workloads shift market share (like GPUs for graphics, ARM for mobile). Cerebras bet on dedicated silicon that doesn't look like a GPU. Wafer-scale chip places SRAM next to compute, solving data-movement bottleneck. Result: 15-18x faster inference. Speed is existential: 'market for slow search is zero.'
24:50 — Investor lessons: hold post-IPO, innovate on lockups: Planet's early investors (Google, Capricorn, Founders Fund) held through 10x run ($5→$50). Gerstner cites MongoDB example: distributed at $3-4B, went to $50B. Cerebras pioneered 'dribble lockup' — shares release over 6 months with performance hurdles; SpaceX adopting similar. Feldman: more absolute dollars made post-IPO. Pendulum swinging back to earlier IPOs.
31:50 — Future convergence: planetary intelligence: Marshall: LLMs are 'blind' without real-world data. Fusing daily Earth observation with AI creates 'large Earth models' / 'planetary intelligence' — sensing + compute in orbit. Feldman agrees on direction, differs on timeline (inter-satellite links remain hard). Both see explosive application growth ahead.
Speakers
Jason Calacanis — Moderator / Host
Brad Gerstner — Founder & CEO, Altimeter Capital (investor)
Will Marshall — Co-founder & CEO, Planet Labs
Andrew Feldman — Founder & CEO, Cerebras Systems