Baiju Bhatt is building a rocket company to put data centers in space — and SpaceX's closed capacity is why

Baiju Bhatt is building a rocket company to put data centers in space — and SpaceX's closed capacity is why

Baiju Bhatt, co-founder of Robinhood and CEO of Aetherflux (which he refers to internally as "Cowboy Space Corporation"), is building orbital data centers and the launch vehicle to put them there. The mission statement he uses: "power humanity from the high frontier."

Bhatt didn't set out to build a rocket. The decision to vertically integrate into launch came from the economics. After working through what it would take to make space-based compute cost-competitive with terrestrial alternatives, the conclusion was that controlling the entire stack — launch vehicle plus orbital data center architecture — was the only path to viable unit economics. SpaceX is the obvious backdrop. Bhatt says SpaceX will absorb most of its own launch capacity, which means anyone who wants to control their destiny in space infrastructure has to build their own way up.

“We announced fundraising and the plans to vertically integrate into launch vehicle plus data centers and space architecture... When you optimize the entire thing — including the launch vehicle, the path to get into data centers in space — that's one of the paths to getting to cost-competitive economics... We're building this thing for our data center and space payload. We're not building it for third-party payloads.”

Launch vehicle design

The design philosophy is deliberately unglamorous. Bhatt wants the Honda Civic of rocket engines, not the Bugatti Veyron — meaning simplicity of architecture and speed to market over peak performance. He's aiming for reusability, but plans to build economics that work on a non-reusable version first, so payloads can start going up before the reusable system is proven.

The first-party payload model changes the risk calculus in an important way. Traditional launch companies built rockets to carry third-party payloads, mostly government contracts, where a failed launch is catastrophic for the customer relationship. Aetherflux is building for its own data centers, which means a failed launch is an internal cost rather than a contract-ending event. Bhatt argues that dynamic pushes toward trying early and often, iterating faster than a company beholden to government payload customers could afford to.

Human spaceflight and third-party launches are explicitly out of scope, at least for now. Aetherflux is a purpose-built data center play that happens to need a rocket.

The compute demand argument

On whether AI hyperscalers and labs are actively seeking non-SpaceX alternatives, Bhatt is measured. He says concerns about dependency on Elon-controlled infrastructure exist, but the dominant force in the market right now is simply raw demand for compute capacity — people are trying to get it wherever they can. The GPU capacity scramble he describes is real: Apple leasing from Google, Google leasing from SpaceX, Meta moving in and out of surplus. Against that backdrop, Bhatt's pitch is that orbital compute is another source of capacity coming online, not primarily a hedge against SpaceX.

A more detailed launch vehicle unveiling — including what the rocket looks like — is coming from Aetherflux soon.