The IPO Comeback: Why Tech Giants Are Finally Going Public | All-In Liquidity IPO Panel
In a Nutshell
Cerebras and Planet Labs went public recently and reported that the IPO process itself changed little about operations—mostly just added cash and public shareholders—while delivering major liquidity and credibility benefits. Planet sees space-based data centers becoming cheaper than terrestrial ones within 2–3 years as launch costs fall, with most AI compute eventually moving to orbit. Cerebras argues purpose-built silicon, not GPUs, is required for real-time AI, delivering 15–18x speedups by placing memory next to a wafer-scale chip.
These notes were generated by AI and may contain inaccuracies.
2026 could be an all-time record for IPOs. Cerebras is described as the AI IPO of the year. Andrew Feldman, founder and CEO of Cerebras Systems, stated the company is participating in something extraordinary and is the fastest bar none. Will Marshall, co-founder and CEO of Planet Labs, discussed how space and AI are a match made in heaven. He compared Planet's work to Google indexing the internet, noting they are indexing the earth and making it searchable. Brad Gerstner, founder and CEO of Altimeter Capital, stated he believes the current wave is the biggest wave in the history of technology and will be incredibly beneficial for America.
The panel features Brad Gerstner, Will Marshall, and Andrew Feldman. Jason Calacanis shared a story about canceling a Japan ski trip to moderate an All-In interview in Davos with Donald Trump after being called by Chamath Palihapitiya. He received a special green badge allowing access through security.
Andrew Feldman is founder and CEO of Cerebras. Will Marshall is founder and CEO of Planet Labs. The panel positions the two companies as gateways to understanding two major trends: AI silicon and space data centers. Discussion begins on the experience of going public, with Cerebras having gone public about three weeks prior and Planet Labs having been public for about a year and a half to two years.
Andrew Feldman noted needing to upgrade his name tag game after seeing the moderator's performance. He described the IPO process as involving an enormous amount of garbage, including meetings with 130 attendees on Zoom and repeated document reviews where only commas move with no value added. He stated that after the IPO event, no additional product was sold and engineering projects made no progress. Feldman emphasized that the core parts of the business remain unchanged, with vendor relationships staying the same whether good or bad. Employees had a party and were excited, but everyone returned to work the next day. The only material change noted was having more money in the bank and new constituents to communicate with.
Brad Gerstner noted he was an investor in Cerebras and on the board a year earlier when they attempted to go public. One investor was the UAE, which created challenges under the prior Biden administration regarding CFIUS. Gerstner observed that everything was hard for nine and a half years, then became easy in the final twelve months. The IPO was priced at $185, above the range which was taken up two times. The stock opened at $320 and was trading at $230 at the time of the panel, representing a $50-60 billion market cap.
Andrew Feldman was asked whether he would advise going public given increased capital, profile, and easier enterprise sales. He stated the answer to timing was getting it wrong for a decade. The company brought employees with more than nine years tenure and their families to the IPO event. Feldman learned engineers owned ties and observed how meaningful the event was for employees and families, describing it as giving external validation similar to a bar mitzvah. One leader whose father was a Chinese immigrant commented that they thought it would have happened faster. After the excitement, everyone returned to work.
Will Marshall discussed Planet Labs' different experience entering public markets. The stock moved from $5 to $50 per share over twelve months, a 10x increase. Planet was one of the first space stocks and initially people had no idea what was happening in space. Marshall stated that going public provides liquidity for early shareholders and cash for the company. He noted it helps the business by giving credibility to customers including agricultural companies, governments, civil governments, and defense and intelligence agencies. Countries dependent on Planet's information care that the company will remain operational. Being public signals the company is here to stay and has access to capital.
Planet operates the largest earth imaging fleet with about 200 satellites that image the entire earth every day. This provides a time series of data rather than static three-year-old images like Google Maps. Applications include agriculture, energy companies, civil governments for flooding and fires, and security. Marshall noted space is part of everyday life through phone communications, GPS, and satellite data. The market has started to understand where space is going. Regarding military versus non-military revenue, security was part of the original plan but represents a larger fraction than initially expected due to current geopolitical needs. The technology enables seeing threats in advance, providing weeks or months of warning to potentially stop conflict. Marshall stated they are not limited to being perceived as a military company and also work with farmers, energy companies, civil governments, and NASA.
Rocket costs have come down four to five times over the last ten years. Satellite miniaturization has occurred, with satellites that previously cost a billion dollars and weighed 20 tons now costing a few kilograms to hundreds of kilograms while performing the same or more functions. This is compared to the mainframe to desktop computer revolution, unlocking new applications. Both reduced launch costs and miniaturization work together.
Will Marshall explained that Planet conducted a study with Google eight or nine years ago comparing costs of terrestrial versus space data centers. The analysis found that when launch costs reach $200-300 per kilogram, it becomes cheaper to put data centers in space. Current costs are just over $1,000 per kilogram, having come down about 10x in the last ten years. With Starship trajectory, this cost point is expected in two to three years. In space, solar panels in sun-synchronous dawn-dusk orbit can collect five times more energy per panel than on the ground with no need for batteries. The infrastructure needed is solar panels, chips, and RF signals. Marshall stated that within ten years most compute will be in space, representing trillions of dollars and larger than other space businesses today. Planet is partnering with Google to launch TPUs into space and has already launched Nvidia GPUs. An early test of Google's TPUs is planned.
Andrew Feldman stated there are still one or two hard problems beyond putting GPUs in space, particularly building clusters in space for communication between chips. He noted the industry is not good at this on the ground and even less so in space. Feldman believes this is an important problem worth attacking and will occur, though possibly on a different timeframe. He compared it to self-driving, where the last 10% proved to be a decade of work. The key enabler is reducing launch vehicle costs to allow experimentation.
Andrew Feldman explained that AI allowed computers to address problems they were previously bad at, including images and language. Starting around 2015-2016, AI opened the aperture for computers to find insight in images and generate and understand language. Simultaneously, vastly more images were being captured terrestrially and via satellites. Feldman stated new workloads present opportunities for market share to change, citing examples of Nvidia's rise with graphics, ARM's dominance in mobile, and new networking companies in the late 1990s. Cerebras made two bets: dedicated silicon would be the answer, and it could not look like a GPU. Feldman argued that to be 20 times better, architecture cannot look the same as existing solutions. The fundamental problem in AI is moving data from memory to compute. Cerebras solved this by building a very large chip the size of a dinner plate with memory right next to compute, using a different type of memory that is vastly faster. When OpenAI uses Cerebras chips, they are 15 to 18 times faster than GPUs, enabling real-time AI responses. Feldman noted there is zero market for slow search or dial-up, and users will not wait for AI.
Planet Labs went public in 2021 via SPAC at a $2 billion valuation. Early investors included Founders Fund, DST Global, and Capricorn. Most early investors stayed in the stock and captured the 10x move from $5 to $50. Google remains the largest single investor and has not sold shares. Capricorn held until recently. This is contrasted with cases where LPs pressure firms to distribute shares early, missing subsequent value creation. Cerebras investors remain under lockup following the recent IPO.
The panel discussed an innovative approach developed with banks called a dribble lockup, where shares are released gradually over six months according to performance hurdles. SpaceX is expected to implement a similar structure. The concept originated years earlier, representing the most innovative lockup approach discussed.
When asked whether VCs should distribute shares immediately upon lockup expiration, Andrew noted that historically more money has been made after IPO than before. Studies consistently show greater returns both in percentage terms and absolute dollars after companies go public. The amount of capital that can be deployed in most venture companies is modest, with only two or three or five outliers allowing significant investment. By the time companies reach public markets, substantially more capital can be put to work if the business is performing well.
The discussion highlighted SpaceX's unique position, noting that much of its value lies in future potential rather than current operations. Most major tech companies went public at valuations of a few billion dollars, not a few trillion. The panel acknowledged that achieving equivalent growth for SpaceX would require aiming at quadrillion-dollar valuations, reflecting Elon's ambitious targets.
Three mega IPOs were referenced as multi-trillion dollar companies where all value accrued to private market investors. Planet Labs was cited as an example where venture capital returns in public markets delivered a 10X outcome. The panel advocated for companies going public earlier, noting that if Andrew had gone public 18 months prior at $10 billion instead of the current $50 billion valuation, that 5X appreciation would have benefited public market investors rather than remaining with private holders.
While some view Anthropic, OpenAI, and SpaceX as the new normal for extended private periods, the panel suggested public markets may be shifting back toward earlier IPOs. Companies in their portfolios are now considering going public at valuations of $1 billion, $3 billion, or $5 billion. After a decade where staying private was heavily promoted, the pendulum appears to be swinging toward companies wanting to go public sooner to compete in public markets like Planet Labs.
The panel emphasized that visible competition creates pressure, but going public earlier brings beneficial scrutiny that sharpens focus. The principle that "steel sharpens steel" and "iron sharpens iron" was invoked to argue that public market discipline improves innovation. Allowing broader participation while subjecting the company to public accountability creates conditions for great outcomes.
The panelists were recognized as testaments to entrepreneurship, with the moderator noting conversations dating back to the beginning of their respective journeys. The next few years were described as particularly exciting due to the convergence of AI and space technologies. Current large language models are powerful but limited because they only absorb text from the internet and lack real-world awareness—they don't understand farm fields, floods, or security situations. Integrating real-world data through planetary sensing and compute systems in space would enable AI to solve actual physical problems, creating what was termed "large earth models" or "planetary intelligence." This convergence is expected to unlock enormous economic opportunities and applications.
Keep All-In Podcast in your library
Save the videos and channels worth coming back to, and find them again in one place.
