HSBC’s central argument is straightforward: the most immediate value of the space economy is showing up on Earth, where it is starting to upgrade existing industries rather than waiting for asteroid mining or Mars settlement to become real businesses.
In a report published on Aug. 18, HSBC mapped that transmission across 12 sectors, including telecoms, semiconductors, power, industrials, agriculture, insurance and finance. The report was led by Raj Sinha, co-head of HSBC Global Equity Research. It said SpaceX has raised $75 billion, and noted that Elon Musk’s compensation is tied to a $7.5 trillion company valuation target and plans linked to Mars colonization.
Starlink has already moved past the scale threshold
As of June 30, 2026, SpaceX’s satellite internet service had 10,200 satellites in orbit and 12 million broadband users across 167 markets. A further 7.4 million monthly active devices were using Direct to Cell service in 30 countries.
HSBC’s telecom analysts said those figures show space-based communications have become infrastructure for multiple industries, no longer just a fallback option for remote regions.
The report pointed to several operating examples. Maersk has installed Starlink on more than 330 container ships. United Airlines plans to retrofit 15 Boeing 737-800 aircraft a month. EpicVue has launched a Starlink package for trucking fleets priced at $99 a month with 250GB of data, covering 99.9% of the U.S. and Canada. In HSBC’s framing, satellite connectivity is shifting from emergency backup to default configuration.
HSBC also said Starlink’s relationship with terrestrial telecom operators is largely complementary. Fiber and 5G still hold cost and performance advantages in dense urban areas, while satellites fill gaps in oceans, deserts, mountain regions and disaster-hit zones where cellular coverage falls short. For telecom operators, the bank said, partnership makes more sense than confrontation. T-Mobile, KDDI and Airtel Africa have all signed Direct to Cell device agreements with Starlink.
Starship is targeting launch costs below $300 per kilogram
HSBC said SpaceX’s core advantage is cost.
It put Falcon 9 low-earth-orbit launch costs at $2,940 per kilogram and Falcon Heavy at $1,520. Starship’s target is $100 to $300 per kilogram. Compared with a historical average of about $18,500 per kilogram from 1970 to 2000, that would mark a decline of more than 95%.
That shift, in the bank’s view, changes the basic assumptions behind the space economy. Material choices in satellite manufacturing, energy plans for orbital data centers and feasibility models for asteroid mining all need to be recalculated.
HSBC’s industrial analysts said about 80% to 85% of SpaceX rocket and spacecraft components are manufactured in-house, giving the company an unusually high level of vertical integration. That leaves less room for outside suppliers and keeps more of the gains from lower launch costs inside the company.
The report gave a more detailed breakdown as well. A V2 Starlink satellite costs about $1 million, with launch costs of $2 million to $3 million. A single Starship launch can carry 60 V3 satellites, each with roughly 1 Tbps of downlink capacity, or 20 times that of V2. Scale and cost are now reinforcing each other.
Orbital data centers remain a backup option for compute
SpaceX plans to deploy 100GW of orbital AI data center capacity by 2040, with test launches starting in 2027.
HSBC estimates orbital data centers currently cost three times as much as ground-based facilities, with convergence only becoming possible around 2035 to 2040. Ground-based data centers remain the first choice. Orbital facilities, in this reading, are more of a strategic reserve. If U.S. grid expansion and nuclear buildout fail to keep pace with compute demand, space offers an alternative route around policy, land and environmental constraints.
HSBC described Terafab as a more immediate concept. Under the report’s description, SpaceX, Tesla and xAI are jointly building a vertically integrated wafer fab covering design, manufacturing, packaging and testing. If that model works, it could challenge the semiconductor industry’s existing horizontal structure. A chain in which Nvidia designs chips, TSMC manufactures them and third parties handle packaging and testing could be replaced by a one-stop approach.
HSBC said the larger risk is not only cost, but excess compute capacity. It estimates annual global AI data center demand additions will stay below 50GW through 2030, while SpaceX alone is planning 100GW of annual additions on the orbital side. If ground operators keep expanding at the current pace, and with HSBC expecting annual capital spending to exceed $1 trillion from 2027, the market could face a compute glut within five years. The report said asset turnover and returns on invested capital in GPU cloud infrastructure are already under pressure, and a SpaceX entry would intensify that trend.
Another effect of overcapacity, HSBC said, would be the commoditization of large language models. If compute is no longer scarce, performance gaps between models narrow. Hardware suppliers would face more pressure, while software and application layers could benefit.
Power, semiconductors and robotics stand out first
Across the 12 sectors covered, HSBC said power, semiconductors and robotics are the three areas being rewritten most directly by the space race.
Power
AI data centers have already pushed annual U.S. power demand growth from 2% to 3% up to 4% to 5%. If the compute race continues, power demand in both the U.S. and Europe could rise 2.8 times by 2050, versus earlier expectations of 60% growth in the U.S. and 80% in Europe. HSBC’s utilities analysts called renewables and grid infrastructure the clearest winners. Nuclear and natural gas carry more uncertainty: ground-based data centers support them, while orbital data centers could become a substitute.
Semiconductors
The semiconductor impact runs in two directions. First, orbital inference would need “low-earth-orbit optimized silicon,” a performance tier between consumer-grade chips and radiation-hardened space chips, creating a new category of demand. Second, if the Terafab model is copied by other large technology groups, traditional foundries could lose orders.
Robotics and industrials
The logic in robotics and industrials is simpler. Asteroid mining depends on highly autonomous robots. Lunar base construction requires heavy equipment. On-orbit servicing needs robotic arms. The report said Caterpillar is already working with NASA on lunar excavation technology, bringing autonomous mining experience from Earth into space applications. Honeywell is supplying navigation systems for NASA’s Artemis missions, and those aerospace technologies can also feed back into aviation and industrial uses on Earth.
Nine other sectors also appear in HSBC’s framework
Beyond those three, HSBC also included telecoms, agriculture, chemicals, metals and mining, healthcare, insurance, finance, transport and logistics, and technology platforms. The bank’s view was that the transmission chain in those sectors is longer, so the near-term effect is less direct.
The report’s broader conclusion is that the space economy does not need to wait for asteroid mining or Mars settlement to lift off. Starlink’s 12 million users, Starship’s cost curve and the potential effect of Terafab on semiconductor industry structure already provide evidence for that shift.
HSBC’s core call is that space is becoming an accelerator for terrestrial industry. Compute demand keeps rising, power needs are climbing, and launch costs are moving toward less than $300 per kilogram at the same time. The first sectors to change, the report argues, are usually the ones with the shortest transmission chain.
The original article said it was a整理与解读 of a third-party broker research report from HSBC dated Aug. 18, 2026, combined with public market information. It also said that ratings, price targets, earnings forecasts and related judgments cited in the piece were the views of the brokerage analysts and represented only their institution’s position, not that of the article’s publisher, and did not constitute investment advice. The piece added that market decisions should be made independently and the content should not be used as a basis for buying or selling securities.
