All-In Podcast: Cerebras CEO Says Going Public Changed Nothing Critical After Decade-Long Journey to $50B Valuation
Panel discussion at All-In Liquidity 2026 event on June 6, 2026
In a surprisingly candid assessment of the IPO process, Cerebras Systems CEO Andrew Feldman told investors at the All-In Liquidity conference that despite a decade of brutal challenges culminating in a blockbuster public offering just three weeks ago, "not a damn thing changes in the important parts of your business." The comment cuts through the typical IPO celebration narrative and offers institutional investors rare insight into what actually matters when evaluating newly public technology companies.
Feldman, whose AI chip company went public at $185 per share before opening at $320 and now trades around $230 with a market cap approaching $50 billion to $60 billion, described the path to public markets as years of "garbage" involving countless meetings with 130 people on Zoom calls reviewing documents where "the commas move and just no value added." The company had tried to go public 18 months earlier but faced regulatory hurdles related to having UAE investors, creating CFIUS complications under the prior administration. According to Brad Gerstner, whose Altimeter Capital was on the board, this meant the company went public at $50 billion rather than the $10 billion valuation it might have achieved earlier, with that 5x appreciation going entirely to private market investors rather than public shareholders.
The Public Markets Thesis: Planet Labs Delivers 10x Returns Post-IPO
The more compelling evidence for early public offerings came from Planet Labs CEO Will Marshall, whose earth imaging company went public via SPAC in 2021 at a $2 billion valuation and has since run from $5 per share to $50, a 10x move that accrued almost entirely to public market investors who held through the journey. "Most of them stayed really well in," Marshall noted, pointing out that Google, the company's largest single investor, hasn't sold a share, while other venture firms like Capricorn only recently began distributions. This stands in stark contrast to the typical venture capital pattern where LPs demand immediate distribution upon IPO, often missing the substantial appreciation that follows.
Marshall's assessment of going public centered on three practical benefits: liquidity for early stakeholders, access to capital, and perhaps most importantly, credibility with large enterprise and government customers. "We work with biggest agricultural customers, big governments, civil governments, defense and intelligence," he explained. "All of those actors want to know you're going to be around and not going to disappear." For companies selling mission-critical infrastructure to sovereigns that become "fully dependent" on the service, being public provides the legitimacy that unlocks contracts.
The Space Computing Thesis: Data Centers in Orbit Within Three Years
Marshall laid out a surprisingly specific timeline for space-based data centers, arguing the economics flip when launch costs reach $200 to $300 per kilogram. Currently at just over $1,000 per kilogram after dropping 10x over the past decade, he expects Starship to hit that threshold "in 2 to 3 years," though he acknowledged "Elon might say it's next week." The core insight is that solar panels in sun-synchronous dawn-dusk orbits collect five times more energy than terrestrial installations without requiring batteries or backup power, fundamentally changing the infrastructure equation for compute.
Planet has already launched some Nvidia GPUs into space and is partnering with Google to launch TPUs for early testing. Marshall believes "no question within 10 years most compute will be putting in space," describing it as "trillions" in market size that will dwarf current space businesses. However, Feldman pushed back on the timeline with characteristic directness, noting "we're not super good yet at building the clusters necessary for the communication between" processors. "We're not good at doing it on the ground. We're really not good at doing it in space," he said, suggesting this may be a problem where "the last 10% is 80% of the time," similar to how self-driving proved to require a full decade to solve the final challenges.
Cerebras Architecture: Why Being 18x Faster Than GPUs Required Dinner Plate-Sized Silicon
Feldman explained that Cerebras was founded on two core architectural bets made in 2015. First, that dedicated silicon would dominate AI workloads. Second, and more critically, that "if you want to be 20 times better than somebody, your architecture can't look like them." This led to the company's signature innovation: a chip the size of a dinner plate rather than a postage stamp, allowing them to place memory directly adjacent to compute and use fundamentally faster memory types.
The results at OpenAI show Cerebras running "15 or 18 times faster than a GPU," which Feldman framed not as a technical benchmark but as a business model unlock. "How big is the market for slow search today?" he asked. "It's zero. How big is the market for dialup? It's zero. You will not wait for AI. We have to deliver it to you in real time." This positions Cerebras not as another GPU competitor but as enabling an entirely different class of real-time AI applications that aren't viable on traditional architectures.
The broader thesis Feldman outlined is that AI opened computing to entirely new problem domains starting around 2015-2016. "We were bad at images for almost the entire history of compute. We could store them and that's about it. We were bad at language. We could store it but that's about it." AI suddenly made computers useful for these previously foreclosed areas while simultaneously generating vastly more image and language data to process, creating what he described as the fundamental driver behind all growth in AI compute, including Nvidia's.
Planet's Military Business and the AI Data Advantage
When pressed on the percentage of revenue coming from defense customers, Marshall acknowledged it's "a bigger fraction today than perhaps we would have guessed" initially, though he declined to provide specific numbers. Planet operates approximately 200 satellites imaging the entire earth daily, creating a time series analysis going back years that's valuable across agriculture, energy, civil government applications like flooding and fires, and security use cases.
Marshall's pitch to investors centered on AI reducing barriers to entry for this data. "AI is only as good as the data it's trained upon," he noted, positioning Planet's daily global imagery as training data for what he calls "large earth models" or "planetary intelligence" rather than just large language models. "All the cool stuff that we're doing with LLMs now is really based on just the text of the internet being absorbed into these models which is incredibly powerful already but they don't know about the real world," he explained. "If you give them real world data, then they can answer real world problems." He estimates the earth observation market alone at $75 billion to $100 billion.
The Lockup Innovation and LP Distribution Dynamics
Gerstner highlighted Cerebras' innovation on lockup structures, describing what he called a "dribble lockup" where shares release over six months according to performance hurdles rather than the traditional single unlock date. He noted SpaceX plans a similar approach. The broader point addressed LP pressure on venture firms to distribute shares immediately upon IPO, which Gerstner argues causes LPs to miss substantial appreciation. He cited a prior investment where his firm bought pre-IPO at $1 billion, distributed at $3 billion to $4 billion per LP demand, then watched the stock run to $50 billion over 24 months while LPs complained about not capturing that appreciation.
Feldman's response was unequivocal: "Historically more money's made after IPO than before. Every single study shows that there is more money to be made both in percentage and in what we care about which is absolute." He noted venture investors can only deploy "relatively little bit of money" in private companies, but by the time companies go public "there's a lot more money there if things are going well and the opportunity to make vastly more is after IPO not before."
Gerstner argued the pendulum is swinging back toward earlier public offerings after a decade where "Andreessen was really pushing stay private forever." He sees companies in Altimeter's portfolio now targeting IPOs at $1 billion to $5 billion rather than waiting for mega-valuations. The three pending mega-IPOs of Anthropic, OpenAI, and SpaceX at multi-trillion dollar valuations represent outliers where "all of that value accrued to private market investors," but Gerstner believes Planet Labs represents the more sustainable model where the 10x appreciation happens in public markets with broader participation.