Grit · 24 Aug 2026 · From the week of 24 August
Data Centers in Orbit by 2028: Baiju Bhatt on His Plans After Robinhood
These are notes on the conversation, checked against its transcript. The episode itself has the full discussion.
In brief
Kleiner Perkins partner Joubin Mirzadegan interviews Baiju Bhatt, co-founder of Robinhood and now solo founder and CEO of Cowboy Space Corporation, which plans to run AI data centers in low Earth orbit inside the upper stages of its own rockets. Bhatt explains how the company started as a space solar power venture (originally called Aetherflex) and pivoted to orbital compute after bottoms-up unit-economics analysis. He argues that the unit economics of the output have to be competitive with terrestrial options, and that this is what pushed him to vertically integrate rockets, satellites and data centers. He also covers the company's ~365 raised to date, his estimate that reaching orbit will take low single-digit billions, his 2028 launch target, and his intense focus on a single use case. He closes with Robinhood lessons: how hard mass-market product-market fit is, his definition of it (beating your own best-case projections), and why a perpetual hiring shortfall is a sign of success, not failure.
For founders
- Bhatt treats unit economics of the output as the single North Star: build a bottoms-up model from parts and manufacturing cost to the cost of the delivered service, and pivot when the model shows a better architecture.
- His Robinhood lesson is that controlling unit economics pushes you toward vertical integration, but each component is still a non-dogmatic build-versus-buy call judged against whether output cost beats the alternative.
- Under an ambitious deadline, Bhatt prioritises by asking which actions will block the timeline if not started now rather than three to six months from now, and he cuts every non-essential 'arts and crafts project' — he wants a 'Honda Civic' rocket engine, not a 'Bugatti Veyron'.
- Bhatt defines product-market fit as consistently beating your own great-case projections (plus long-term retention for consumer products), and warns that many startups are already dead without realising it.
- A perpetual shortage of people to keep up with demand, which was Robinhood's chronic thorniest problem, is in Bhatt's and Mirzadegan's view a symptom of product-market fit worth wanting.
For revenue leaders
- Bhatt's working test for product-market fit is performance beating your 'great expectations' scenario, not merely hitting plan; for consumer products he pairs it with long-term retention.
- Robinhood broke through against entrenched brokers by combining insight into how people used mobile technology with systematic customer conversations that fed directly into product development.
- When recruiting a hesitant senior candidate, Mirzadegan spoke directly to her spouse and framed the startup as having 'a supply problem, not a demand problem.'
What was said 25, most useful first
Layering a higher-value service such as compute on top of a commoditized input such as energy creates more margin to reach break-even. Listen
Bhatt framed the strategic question as 'space solar power plus something else'. Selling raw energy is a truly commoditized service. Adding compute on top inherently leaves more margin to hit break-even numbers. That reasoning produced the formula of space solar power plus compute, minus power beaming.
“if you take a truly commoditized service like energy and you layer stuff on top of it, meaning compute for example, that inherently there's gonna be more margin to work with there to kind of get to these break even numbers.”
Designing the rocket, satellite and data center as one integrated system led Cowboy Space to make the rocket's upper stage itself the data center. Listen
Bhatt gives two realisations. First, two GPU racks are more useful next to each other than apart, so one big satellite beats two small ones. Second, the thorniest part of a space data center is dissipating heat, which normally means carting radiator mass to orbit. Instead the company reuses the rocket's own metal: in orbit the nose pops off, solar panels deploy, and the fairing's cylinder folds out as a heat-dissipating radiator, eliminating a separate satellite and, the company believes, giving one of the lowest-cost ways to do compute from orbit.
“the upper stage of the rocket is itself the data center. There's no satellite, the rocket becomes the data center satellite once in orbit.”
Cowboy Space builds its rocket program for exactly one use case and aims for a 'Honda Civic' rocket engine rather than a 'Bugatti Veyron'. Listen
The rocket program will not do human spaceflight or carry other customers' payloads, though Bhatt allows it may do so someday. He says it needs a rocket that works, not the world's greatest rocket. Without 25 years as a rocket company, Cowboy's way to win is to be surgical: cut 'arts and crafts projects', pick what drives the timeline, and say no to a lot.
“That means cutting out all the arts and crafts projects and being ultra, ultra, ultra judicious with picking the things that drive our ability to do this and frankly saying no to a lot of other things.”
Bhatt defines product-market fit as consistently beating your best-case projections, and for consumer products as strong long-term retention. Listen
He frames forecasts as mediocre, good and great expectations. If you are beating even the great case, something resonates with users more than you anticipated. For consumer products he adds the test of whether people keep using the product over time or whether usage dwindles to zero.
“If you're doing better than great expectations, like you generally have product market fit, like something is going on that is more resonant with the people that are using your product than what you thought.”
Many operating startups are already dead and their founders haven't come to grips with it, because product-market fit is so elusive. Listen
Drawing on several unsuccessful startups he built before Robinhood, Bhatt describes founders convincing themselves the company will be a thing when it may already be a 'zombie startup'. He would much rather face the pains of chasing the market, such as being understaffed, than be in that position.
“a lot of the times when you're operating a startup, you're kind of dead and you don't know it.”
Mirzadegan courted a hesitant head-of-design candidate by talking to her husband and pitching the startup as having a supply problem, not a demand problem. Listen
For a company he incubated at Kleiner Perkins, Mirzadegan judged that leaving a cushy job for a startup was a personal, household decision, so he spoke directly with the candidate's husband. When the husband asked when the supply problem would be fixed, Mirzadegan said he hoped never, because solving it would mean they were no longer chasing the market. He described current competition for talent against AI labs and overfunded startups as 'full hand-to-hand combat'.
“I said, Hey, we have a supply problem, not a demand problem.”
Bhatt's founding theses at both Robinhood and Cowboy Space rest on the idea that emerging platforms spawn entire economic verticals native to them. Listen
At Robinhood in 2012 his view was that mobile was an emerging platform. Beyond apps like maps and photo sharing, he believed whole verticals such as investing and financial services would exist natively on it, not just as apps built by legacy firms. At Cowboy Space he applies the same lens to low Earth orbit: beyond existing verticals like earth imaging, defense and government, he expects energy to be one of the next industries native to that platform.
“if you think of mobile as a platform, that there were gonna be verticals of the economy that natively existed on this platform.”
Bhatt picks space business models with bottoms-up cost analysis aimed at one question: can the service delivered from space be cost-competitive with doing it on Earth? Listen
His central question was how to make an output service from space cost-competitive with the terrestrial alternative, comparing for example space internet with fiber, or space energy with ground generation. He modelled satellite designs from parts cost and manufacturing cost up to the economics of the output service. For orbital compute he breaks the cost of a GPU hour down to the cost and quantity of solar panels, radiators and other components.
“doing bottoms up economic analyses where if you build a satellite that does these things, what are the parts cost? what are the manufacturing costs add up to? And in turn, what does that tell you about what the services, output economics are going to be?”
Cowboy Space pivoted from beaming solar power to Earth to orbital data centers because using the power in space avoids the energy lost in transmission and conversion. Listen
The company began as a space solar power company called Aetherflex, seeded by Bhatt. Because data centers were becoming where energy demand was heading, he started looking at powering them. He realised that beaming power down loses a lot of energy converting from electrical to optical and back. Harnessing and consuming the power in orbit should therefore, from first principles, produce better economics.
“If you could just harness the power in space and use it in space, then you don't have all these transmission losses.”
Bhatt's Robinhood lesson is that controlling the unit economics of your output necessarily leads to vertical integration, though each component remains a non-dogmatic build-versus-buy decision. Listen
He says Robinhood vertically integrated a lot over its history to control unit economics. Cowboy Space is building much of its stack, including rockets, though not all of it. Some component technology can be bought, and each case is decided on whether it keeps the output cost competitive with the terrestrial option.
“if you want to control the unit economics of your output service, it necessarily leads you to vertically integrating, right? To controlling the sort of components that drive it.”
There is nowhere near enough launch capacity for demand, so Cowboy Space concluded it must build its own launch to control its scale and economics. Listen
Asked why Cowboy can't just use existing launch providers, Bhatt said there isn't enough mass-to-orbit or rideshare capacity available. He noted that SpaceX has said it will use much of its own payload for its own projects. To control how it operates, its scale and its unit economics, Cowboy concluded it had to treat launch as part of a vertically integrated system.
“there is nowhere near enough launch capacity that's available for what everybody in the world wants to do.”
To hit an ambitious 2028 launch target, Bhatt prioritises whatever will block the timeline if work doesn't start now rather than three to six months from now. Listen
Cowboy has set a goal of launching its first in-house built system at the end of 2028. Bhatt says that is within, or shorter than, the time it takes to build a large terrestrial data center. He runs a mental algorithm for which actions must start now to protect that date. That is why much of his effort in recent months went into standing up a rocket engineering team in Seattle.
“what are the things that are gonna prevent that timeline from happening? If we don't take specific action to like, get some ball rolling now versus three, six months from now, prioritizing accordingly.”
Cowboy Space has raised about 365 in total, including a 275 Series B, and Bhatt expects reaching orbit to take low single-digit billions. Listen
Bhatt gave the figures as a 275 Series B on top of about 90 raised earlier, for roughly 365 total; he did not state units, though the context implies millions. The company has roughly 50 to 60 employees and is growing quickly. He expects the 2028 launch to require low single-digit billions, so fundraising will be a big part of his job.
“We announced 275 Series B. Before that, we'd raised about 90. So total, you know, capital raised is around 365.”
Bhatt seeded Cowboy Space with his own money, reasoning that committing his time without his capital didn't make sense. Listen
He was the sole seed investor when the company was a space solar power startup. He explained that he weighs time against resources: his time is extremely valuable, and at this stage of life he likely has less time than money. Investing his time but not his money 'doesn't compute' for him.
“I think about the balance of time versus resources. I think my time is extraordinarily valuable.”
At a capital-intensive hardware startup, Bhatt's CEO job has become mainly fundraising and business development. Listen
Bhatt joked that he ended up with the 'emails and spreadsheets job' while engineers build hardware in the clean room. He said fundraising is an important input, but there is also a large amount of business development.
“I don't know how this happened, but I got stuck with the emails and spreadsheets job at this company.”
The thorniest unsolved problems inevitably land on the founder's desk, including mundane ones like an empty HR function. Listen
Bhatt describes founders being served problems that the organisation isn't mature enough to solve or that nobody knows how to handle. At Robinhood, HR turnover once left the company with nobody in HR for a summer while engineers were upset, and Bhatt had to act as the HR person until a hire was made. If someone else could solve a problem, it would never reach the founder.
“Like the worst the thorniest problems end up on your desk as the founder of the company”
Founders can't work nine-to-five, but they don't need constant grind either; they must pick the right moments to push hard. Listen
Bhatt says entrepreneurs have to exert real force to bring an idea to life. He does not think that means 'grind maxing' every night and weekend. Instead he times intense pushes, and like Michael Jordan he deliberately looks for things to motivate himself, such as using his long-held love of science as fuel.
“You can't clock in a nine to five and expect to be successful as an entrepreneur.”
Bhatt credits Robinhood's breakthrough against entrenched brokers to insight into how people would use technology plus systematic customer conversations feeding product development. Listen
He calls mass-market consumer product-market fit extraordinarily difficult, especially when many talented, well-resourced incumbents already serve the market. Robinhood broke through by pairing its view of how people would use mobile with systematic processes for understanding customer needs and flowing them into what it built.
“through pretty systematic ways of talking to customers and understanding what their needs are, and, you know, flowing those into the product development.”
Building a hardware company forces geography-specific hiring and facilities: you go where the specialised talent is. Listen
Unlike Robinhood, a digital product with no factories, Cowboy Space needs a rocket engine factory in Seattle, a vehicle manufacturing site that is still TBD, and launch pad infrastructure. The rocket and satellite engineering teams are in Seattle, while the team building the first two satellite missions is in the Bay Area. Long term he expects computer science and engineering, many executive functions and probably compute sales to sit in the Bay Area, and notes that people who build launch towers are concentrated in places like Florida.
“you need to hire people that know how to build rocket engines. and you have to kind of go where the talent is.”
Bhatt expects AI agents working in the background for billions of people to require a tremendous amount of compute, and says compute is already constrained today. Listen
He calls this a pedestrian base case that doesn't require believing in humanoid robots. As evidence of current scarcity, he says Claude Code times out on him frequently because the providers are running out of compute, and the same goes for OpenAI and Gemini.
“what about the world in which you know billions of people around the world have AI agents that are working in the background for them like doing their menial shit for them like that's gonna require a tremendous amount of compute”
The window for space data centers is time-limited. Listen
Mirzadegan noted that time, alongside unit economics, is a key variable given the current race to build terrestrial data centers. Bhatt agreed the opportunity won't last forever. That view underpins the company's aggressive end-of-2028 target for its first in-house built system.
“I don't think this opportunity is going to exist forever.”
Robinhood's most persistent thorny problem was hiring enough engineers, which Bhatt sees as a pain born of success. Listen
Bhatt said that through most of Robinhood's scaling the company was understaffed and under-resourced at every step, which was very stressful. Mirzadegan framed this as the gift of product-market fit: demand outrunning capacity. Bhatt agreed that many of the toughest problems come from something going well.
“a lot of like the really tough things are in some way, shape or form, born of a thing that went well as well, right?”
Space becoming fashionable has helped Cowboy Space recruit, after years in which the sector was unfashionable. Listen
Bhatt says space was not hyped when he started about two years ago; people asked why he wasn't building an enterprise finance SaaS company. He says the rise of commercial space as a consensus view has inevitably helped hiring. He noted that 10 to 15 years ago the U.S. launch industry seemed to be in decline.
“Space was not hypey at all when I started doing this.”
Solo founding as an experienced operator is far lonelier, and Bhatt says he didn't consciously consider finding a co-founder this time. Listen
After roughly 14 years co-founding Robinhood with Vlad, Bhatt is building Cowboy Space alone. He says he felt a strong personal pull to build a space company and wanted to try it himself. He adds that finding a co-founder is harder at his stage, being older and used to operating businesses.
“It's a lot lonelier, that's for sure.”
The people who win are the ones who want it more, because they don't give up. Listen
Mirzadegan quoted Spotify's Daniel Ek: whatever other reasons Spotify gave for existing alongside YouTube and Apple, the truth was it would care more about music than anyone else. Bhatt agreed and generalised the point as a lesson he relearns repeatedly.
“The people who get it in life are the people who want it more because they just don't give up.”