Starcloud Says It Will Mine Bitcoin From Orbit. The Physics Hasn't Signed Off.

The press release is four hundred words long and contains no numbers. No kilowatts. No kilograms. No dollars. No date. A company called Starcloud says it will place a satellite in orbit and mine Bitcoin with it, and that this will "revolutionize energy efficiency" and "reshape the economics of the industry." That is the entire payload of information. I have been reading crypto press releases for nineteen years, and I have learned that the words a document avoids are usually more eloquent than the words it contains. Tracing the ghost in the machine means reading the shape of the absence. Here, the absence has a very specific silhouette: thermal. Somewhere between "solar power in space" and "reshape the industry," there is a radiator that nobody wants to discuss. So let's discuss it. The lineage Space and crypto have been flirting for a decade. Blockstream beamed the Bitcoin blockchain down from geostationary orbit in 2017 — a broadcast, not a miner, although the distinction rarely survived the headlines. Rideshare payloads have carried hashboards and node hardware past the Kármán line more than once. Each time, the narrative burned bright for a quarter and then went dark, and each time the industry forgot it had happened before. What returned in 2024 and 2025 is not new. It is old wood, rekindled by a genuinely new fire: AI's appetite for compute has made "orbital data center" a phrase that venture capital can say out loud without embarrassment again. I spent early 2024 with a small group of legacy finance people dissecting the BlackRock spot filing, and the lesson I took from that room still holds. Institutions do not buy technology. They buy legibility. Space is legible. Mining is legible. Put them in the same sentence and capital listens before it reads. And Starcloud is, notably, a company. Not a protocol. Not a token. There is no governance forum, no unlock schedule, no treasury, no emissions curve. I want to be precise about this, because it changes everything about who is exposed and how. There is no subsidized liquidity flywheel to unwind, no vesting cliff waiting to dump on you, no governance oligarchy to capture. What replaces those familiar risks is an older, quieter set: an opaque private capital structure, no public market, and no way for you or me to perform diligence on anything except a press release and a company name. That is the trade. Fewer games, more silence. For someone who has spent years reading the silence between the blocks, that is not necessarily a complaint. But it is not a comfort either. This is a bear market, and in a bear market the question is never whether a story is beautiful. The question is which balance sheet survives the next eighteen months. Where the energy actually goes The claim has three stages. The press release mentions only the first. The solar constant at Earth's orbital distance is roughly 1,361 watts per square meter. No atmosphere to attenuate it, no clouds, no weather, no night — except there is night. A satellite in low Earth orbit circles the planet every ninety minutes and spends thirty to forty-five of those minutes inside Earth's shadow. Realistic duty cycle lands between sixty and seventy percent. The "twenty-four-hour sunlight" story only holds in a dawn-dusk sun-synchronous orbit, which is a narrow and increasingly crowded slice of sky. Stage one is therefore already a qualified claim, not a headline. Stage two is the part that gets deleted from the pitch deck. Vacuum has no convection. There is no air to carry heat away from a die that is doing nothing but producing heat. Every watt a miner consumes as computation becomes a watt that must leave the spacecraft, and the only exit is radiation. Radiation is governed by the Stefan-Boltzmann relation: emitted power scales with the fourth power of absolute temperature. At 300 Kelvin, a blackbody surface radiates about 460 watts per square meter. Allow a two-sided radiator with realistic emissivity and you are near 900. An S21-class ASIC draws roughly 3.5 kilowatts. Do the division and you need on the order of four square meters of radiator — before margins, before the sun-facing side is accounted for, before fifteen years of surface degradation. Then size the solar array for that same machine, add a battery to carry it through eclipse, add attitude control and structure and a downlink, and you arrive at a platform whose mass and volume dwarf the thing it exists to power. The satellite is not a container for a miner. The miner is a small, inconvenient passenger aboard a radiator with wings. There is a second engineering cost that never makes it into the narrative: upgrade cadence. Ground mining economics are ruthless about silicon. When a more efficient generation ships, operators swap boards and write down the old ones. You cannot swap a board in orbit. Launch windows slip by quarters as a matter of routine; by the time a hashboard reaches its slot in the sky, it may already be two generations behind the machines it is competing against in Sichuan and Texas. Space mining offers a fix to the energy line item and then breaks the capital replacement line item entirely. Stage three is radiation, and here — unexpectedly — mining wins. Cosmic rays flip bits. Single-event upsets corrupt computation, and in a satellite performing navigation or imaging that corruption is silent and dangerous. Bitcoin mining has a structural property most compute workloads do not: the work is self-verifying and deliberately wasteful. A miner proposes candidate hashes; the network checks them cheaply; wrong answers are discarded without consequence. A flipped bit wastes a share. It does not corrupt a ledger, because the ledger is not being written by the machine that flipped. This is the one technical point the narrative gets accidentally right, and I have not seen anyone say it out loud. SHA-256 mining is unusually fault-tolerant — not because the hardware is robust, but because the protocol throws away almost everything it computes by design. If orbital compute has a natural first customer, it is a workload whose error rate is priced in from the beginning. The code remembers what the market forgets: not every failure is a catastrophe. Some failures were always expected. The only number that matters None of the above decides whether Starcloud works. One variable does. Launch cost per kilogram. Today, rideshare to low Earth orbit runs somewhere in the $1,500 to $6,000 per kilogram band depending on orbit, provider, and schedule. At Starship-class reusability targets — the hundred-dollars-per-kilogram conversation — the equation inverts and orbital mining stops being absurd. Between those two numbers lies the entire business. Everything else is commentary. Which is why the phrase "energy efficiency" in that release is doing so much dishonest work. Space offers cheap marginal energy. It offers nothing on total cost of ownership. You pay for that energy up front, in kilograms, at the launch pad, long before the first hash. Framing a capital expenditure problem as an electricity price problem is not a technical claim. It is narrative engineering, and it is skilled narrative engineering — I withdrew to Patagonia for three months after the Terra collapse precisely because I had watched how persuasive that kind of framing becomes when enough people agree to believe it. The quiet ruin when the algorithm broke was never arithmetic. It was a story everyone had already signed. The contrarian read Take the company at its own name. Starcloud. A data center company. If that is true, mining is not the business. Mining is the bridge. Consider what an orbital compute provider needs before it can sell AI inference to anyone: customers, bandwidth, latency guarantees, thermal headroom, and a power budget that survives contact with reality. Now consider what Bitcoin mining needs. An intermittent supply it can throttle without penalty. A network that tolerates hundreds of milliseconds of latency. A hard-asset payout in a currency no counterparty can freeze. Verification performed by strangers, for free, forever. Mining is the only workload that fits inside that envelope. It is the minimum viable cash flow — the payload that pays rent while the actual thesis gets built. Whether Starcloud says so or not, hashboards are a revenue bridge to an orbital data center argument, not the destination itself. The mining is not the ambition. It is the bootstrap. Two consequences follow, and both are more interesting than the launch. First, the biggest winner of a successful orbital mining demonstration is not the miner. It is the launch provider. Every kilogram of radiator and solar array that a mining satellite requires is a new, recurring, non-constellation order book, paid in advance. That is where the durable value accrues, and it is the position least exposed to whether the mining itself ever turns a profit. Second, and stranger: nobody has resolved which jurisdiction taxes Bitcoin mined in orbit. Not in national airspace, not on the high seas — in a legal vacuum above the atmosphere, governed by the 1967 Outer Space Treaty's supervision regime, with no precedent for where the coin is "earned." That is a case waiting to be filed, and the first filing will be ugly. What to watch Four signals, ordered by diagnostic value. Whether the payload reaches orbit and powers on at all — a binary event that press releases cannot substitute for. Whether the company ever publishes an in-orbit hashrate in terahash, rather than in adjectives. Whether a cost-per-Bitcoin figure appears anywhere near a ground-mining average, because that single number decides whether this is a business or a demonstration. And whether the Starship price curve bends below five hundred dollars per kilogram, which would rewrite the question for every operator on Earth, not just the one in orbit. When the herd wakes, the signal has already faded — so watch the numbers, not the launch. The launch is a photograph. The radiator is the truth.