SpaceX aims to put Nvidia's top-end AI accelerators into orbit by 2027, a schedule Elon Musk says he is "highly confident" the company can hit — and the first credible attempt to move AI compute off a power-constrained planet. SpaceX shares closed at $152.71 on Sept. 18, down 1.36%, against a 52-week range of $104.83 to $225.64.
The plan adapts satellites SpaceX already builds for Starlink, replacing communications payloads with Nvidia computing hardware and adding solar arrays to power them. Nvidia has been named the exclusive supplier of the chips, and it already holds a sizable stake in SpaceX — an ownership position that makes its exposure to the outcome deeper than a typical vendor relationship.
"The physics of space are unforgiving, but the economics of terrestrial power are getting worse every quarter," Musk said, describing an initial launch late next year with larger deployments following in 2028. He has floated a long-run ambition of up to one million compute-carrying satellites.
The cooling problem is the whole ballgame
The obstacle that matters most is heat. In a vacuum there is no air to carry warmth away, so an orbital data center must reject waste heat by radiation alone. One estimate cited in coverage of the plan puts the radiator requirement for a single orbital facility at millions of square feet — a structure far larger than the computing payload it would protect. Terrestrial hyperscalers solve the same problem with evaporative cooling and river water; neither exists at 550 kilometers.
Radiation shielding, the near-impossibility of sending a technician to repair a failed unit, and the cost of moving petabytes between orbit and ground add to the list. None is disqualifying on its own. Together they explain why even bullish analysts do not expect meaningful revenue for years, with some penciling in the first serious orbital capacity no earlier than 2029.
For scale, Nvidia's data center segment generated the overwhelming majority of its revenue in its most recent quarter, and the company's entire growth story rests on terrestrial buildouts from Microsoft, Amazon, Alphabet and Meta. Orbital capacity in the 2027-2028 window would be a rounding error against that base — a feasibility test, not a demand driver.
What the 2027 date is actually worth to SPCX and NVDA
The near-term case for both stocks remains earthbound. SpaceX's revenue comes from launch services and Starlink subscriptions; Nvidia's from the ongoing AI infrastructure boom. SpaceX carries a Moderate Buy consensus rating with a $221.42 price target, implying roughly 45% upside from the Sept. 18 close — a gap that reflects the market's uncertainty about how to value a company that is simultaneously a launch provider, a satellite operator and now a prospective compute host.
SpaceX shares have risen about 45% since the start of August, while Nvidia has added roughly 10% over the same stretch. Both are close to flat over the past month, which is precisely why a new growth narrative carries outsized weight: neither has an obvious near-term catalyst left in its existing business.
The competitive read matters too. If orbital compute works, SpaceX owns the launch vehicle, the satellite bus and the network — assets already paid for — and could add a business at a marginal cost no rival starting from scratch could match. That is a structural advantage over Amazon's Project Kuiper and any terrestrial operator that would need to buy its way into orbit. Nvidia's position is simpler and arguably larger: exclusive chip supplier to a market that does not yet exist, on top of the one that does.
The honest framing is that this is a catalyst to track, not to underwrite. The 2027 date is a feasibility milestone. Watch for a named payload on a published launch manifest, a confirmed Nvidia hardware specification for space-rated operation, and any statement from either company that moves the date. Until a manifest entry exists, the orbital data center story is a call option on a launchpad — valuable, but not yet revenue.
This article is for informational purposes only and does not constitute investment advice.