Two arXiv Papers Model Launch Pricing and Orbital Computing — Neither Compares Indian Rockets With SpaceX

The research captured for this topic consists of two arXiv working papers. One develops a model of competition and vertical integration in orbital launch markets; the other evaluates whether large-scale AI data centers in low Earth orbit could be a cost-effective alternative to terrestrial facilities. Both are first-version submissions, and neither is a news event with a verifiable reporting date in the capture. ^{[3] [3] [2] [2]}

The launch-market paper by Akhil Rao reports that U.S. orbital launches grew roughly tenfold over 15 years, with about three-quarters of them on SpaceX's Falcon 9 and roughly three-fifths of those Falcon launches deploying Starlink satellites. A Wright's law calculation, which the author describes as back-of-envelope, suggests real Falcon 9 launch costs should have fallen about 70 percent from 2012 to 2026, while advertised prices fell less than 6 percent in real terms. The paper's model attributes that gap to vertical integration between a launcher and its captive constellation, with cost reductions captured as capacity rent rather than passed through to external buyers. ^{[3] [3] [3]}

The orbital-compute paper by Kees van Berkel compares orbital and ground-based systems across launch cost, power generation, cooling, radiation exposure, atmospheric reentry and compute-network performance. It cites an expected Starship launch-to-LEO cost of about $200 per kilogram, and calculates that launching 1 GW of AI compute would involve roughly 10,000 tons of satellites at a cost of about $2 billion on that basis. For comparison, the paper puts a terrestrial 1 GW AI data center at $35–60 billion to construct and about $1 billion a year to operate. ^{[5] [5] [5] [5]}

Van Berkel's modelling concludes that while LEO-based inference may be feasible, training frontier-scale large language models in orbit is unlikely to be competitive with terrestrial data centers, a finding tied to the shift from terrestrial Clos networks to space-based mesh networks using laser inter-satellite links. The paper's cost figures rest on cited third-party estimates and on an attributed claim about satellite mass, not on independent measurement. ^{[4] [4] [5] [5]}

Neither paper examines Indian launch vehicles or presents a per-kilogram comparison between India and SpaceX. A DRDO weekly defence-news digest lists a news article whose headline asks what ISRO can do to reclaim a cost advantage as SpaceX offers some of the cheapest satellite launches, but the digest carries only the title and source line, with no figures or methodology. A captured arXiv author-search page for van Berkel lists his other work in logic, survey methodology and temporal reasoning, with no Indian launch-cost study. ^{[1] [4] [4]}

The commissioning question about a Cambridge study's $13,302-per-kilogram figure for Indian launch to low Earth orbit cannot be checked against the captured material: no such study appears, and no passage states that number or decomposes the gap into cadence, vehicle design, propellant, range costs or institutional overhead. For a usable comparison, readers would need the primary study's methodology and its stated treatment of marginal versus fully loaded costs. ^{[2] [4] [1]}