The satellite that intends to mine the first Bitcoin in space — late 2026. ASIC miners, an Nvidia Blackwell GPU and an AWS server blade on one 100×-power spacecraft. This page holds the claims to the light now, and switches to live tracking the day it has an orbit.
Starcloud-2 is the spacecraft the company’s $170 million Series A — raised to build solar-powered orbital data centres — is primarily paying for. Public statements and investor coverage describe a machine generating roughly one hundred times the power of Starcloud-1, deploying what the company calls the largest radiator ever flown — because waste heat, not electricity, is the binding constraint on orbital compute — and carrying three distinct payloads: a GPU cluster including an Nvidia Blackwell chip, an AWS server blade running commercial workloads for partners including AWS, Google Cloud, Nvidia and Crusoe, and a set of Bitcoin-mining ASICs.
The ASICs are the headline. CEO Philip Johnston’s framing, in a March 2026 interview and on X: Starcloud-2 “will be the first to mine Bitcoin in space.” His economic case is that mining hardware is radically cheaper per kilowatt than AI hardware — roughly $1,000 per kW for ASICs against tens of thousands for a top GPU — so orbital solar power pointed at SHA-256 makes a cheap first market.
Two versions of the Bitcoin story exist in the company’s own materials, and they are not the same size.
The public version, from the CEO: Bitcoin mining draws on the order of 20 GW continuously on Earth, it makes no sense to keep doing that terrestrially, and in the end state it all moves to orbit. That is the version the crypto press ran with.
“We think we’ll be the first to mine a coin in space.”
Philip Johnston, CEO of Starcloud — March 2026
The investor version, from Series A coverage: the ASICs are a revenue bridge — a way to monetise idle capacity when AI utilisation is low. A bridge is a sensible engineering hedge. It is also a much smaller claim than the relocation of a 20 GW industry. Both statements are the company’s own; which one Starcloud-2’s actual duty cycle resembles will be measurable in its downlink behaviour and any mined blocks.
A transparent back-of-envelope, assumptions stated. If Starcloud-2’s power budget is on the order of 100 kW and every watt went to modern ASICs at roughly 15–20 joules per terahash, the spacecraft would produce about 5–7 PH/s. The Bitcoin network runs on the order of a million PH/s. One satellite is therefore in the region of five parts per million of network hash rate — a statistically expected yield of a fraction of one Bitcoin per year, before radiation, thermal and bandwidth realities take their cut.
That is not a debunk; it is the scale. The play is not this satellite’s coins — it is demonstrating the cost curve that would justify the company’s FCC filing for up to 88,000 satellites. One spacecraft mining a token amount proves plumbing. The economics only exist at constellation scale, which is why the filing and this launch are the same story.
Starcloud’s energy case rests on near-continuous sunlight, which in low Earth orbit means a dawn–dusk sun-synchronous orbit near 97° inclination. Starcloud-1 did not get one: it flew a 45.4° rideshare orbit that spends roughly 38% of every revolution in Earth’s shadow, running on battery — a demonstrator that does not demonstrate the pitch’s central assumption.
So the single most informative number on launch day is Starcloud-2’s inclination. Near 97°: the company paid for the orbit its economics require. Another mid-inclination rideshare: the continuous-solar story remains untested at 100× the power, and the mining arithmetic above degrades by every eclipse minute. The element set settles it within hours of catalogue entry, and this page will publish it.
In July 2026 the five-digit satellite catalogue ran out at 69,999, and newly catalogued objects now receive six-digit numbers that the legacy TLE format cannot carry natively. Starcloud-2 will almost certainly be catalogued in the new range — which quietly breaks any tracker still parsing classic TLEs. This site’s data pipeline already speaks the successor formats, so tracking here starts the day the catalogue entry exists.
Starcloud-1 established that a space data center in miniature survives launch and runs in orbit: an H100 executing real workloads, including training a language model in space. It did not establish the energy model (wrong orbit), the disposal model (its deorbit clock runs out around 2 October 2026 with no passive backstop from 501 km), or the economics. Starcloud-2 inherits all three open questions at a hundred times the power. Our Starcloud-1 page tracks the first two live.