Behind the video
88,000 Satellites?! Starcloud CEO Interview (Philip Johnston) 🛰️
Overall Thesis
This is an interview with Starcloud's CEO about building orbital AI data centers, covering the company's satellite launches and the broader rush toward space-based compute by SpaceX, Google and Amazon. Tesla and Nvidia are discussed mainly as hardware/chip suppliers and speculative business angles rather than as direct stock calls, while the host offers a brief but genuine bullish aside on Bitcoin's potential to be mined in space.
Narratives
NVDANVIDIANvidia H100 GPUs are the hardware currently powering Starcloud's orbital data centers, and the host jokingly raises the idea of a $50 trillion Nvidia valuation if compute demand keeps growing, but immediately disclaims it as not a real prediction.
Key Arguments
- Nvidia H100 chips are the GPUs flown on Starcloud 1 and used for testing on the ground
- Host suggests compute demand growth could be enormous for chip suppliers like Nvidia
Risks acknowledged
- Host explicitly says he doesn't want to make 'crazy predictions' like a $50 trillion valuation, undercutting the claim as a genuine forecast
TSLATeslaThe guest speculates that Tesla's new chip fab is being built partly to produce a space-optimized AI inference chip, which Starcloud would be interested in using if Tesla ever sold it externally. No view on Tesla's stock price or investment return is expressed by either speaker.
Key Arguments
- Guest believes one of the main reasons Tesla is building its fab is to produce a space inference chip
- Starcloud would happily use a Tesla-designed space chip since it could offer mass and cost advantages over adapting Nvidia hardware
Risks acknowledged
- Guest notes Tesla currently doesn't sell any of its AI chips, so it's uncertain whether they would sell this one externally
BTCBitcoinThe host argues that the ability to mine Bitcoin in space using cheap ASICs is a compelling and bullish new use case for orbital compute, potentially becoming a whole new industry, though the idea remains speculative and unquantified.
Key Arguments
- Bitcoin's value already derives from energy consumption and decentralization, and space mining could add a new scalable avenue for that
- ASICs are about 30x cheaper per watt than GPUs, making space-based Bitcoin mining potentially more economical than GPU-based compute
- Starcloud plans to fly Bitcoin mining ASICs on its second spacecraft and expects to be first to mine a coin in space
Risks acknowledged
- ASICs have a shorter useful half-life than GPUs, which is a drawback since the orbital infrastructure dies with the chip
- Terrestrial Bitcoin miners can access much cheaper energy (2-3 cents/kWh) than the power achievable in space-grade data centers (8-9 cents/kWh), narrowing the cost advantage
Predictions (1)
Hedges & Caveats
- Host explicitly disclaims making a 'crazy prediction' when floating a $50 trillion Nvidia valuation
- Guest's claim that Tesla will build a space inference chip is his own speculation, not confirmed by Tesla
- Bitcoin-in-space bullishness is framed as a new, unproven use case ('could be a whole totally new industry')