Chanos says space data-center costs should make SpaceX investors cautious

Google’s reported view on space data centers

TIKR.com reports that the weekend debate over SpaceX’s future centered on orbital compute, after The Information said Amanda Peterson Corio, who leads data center energy at Alphabet’s Google, placed the cost of building data centers in space at as much as 100 times the cost of similar facilities on Earth. According to the report, she also said that U.S. power supply is enough and that permits are the slower part of the process. TIKR.com quotes her saying, “Let’s explore all the terrestrial options before we go extraterrestrial.”

Chanos says the cost gap should slow modeling assumptions

Short seller Jim Chanos took that reported estimate as meaning roughly $5 trillion per gigawatt of compute, according to TIKR.com. He wrote on X that the difference between Google’s estimate and SpaceX management or bull assumptions should make investors hesitate before trying to model the business. TIKR.com says Chanos also argued that the economic gains from AI compute would ultimately go to GPU makers such as NVIDIA rather than to the companies renting out the hardware or facilities.

SpaceX revenue estimates and the orbital compute debate

TIKR.com says it tracks 39 analysts covering SpaceX. The article notes that analysts expect SpaceX’s revenue to increase from $48.9 billion this year to $336 billion in 2028. Elon Musk said on X that orbital compute deployed per month probably exceeds terrestrial compute near the end of 2028. But Sequoia Capital partner Shaun Maguire, whose SpaceX model helped spark the weekend debate, said orbital compute would not be the company’s dominant revenue line until 2030 or 2031.

Starship, launch cadence, and the scale of the buildout

According to TIKR.com, Maguire says SpaceX needs Starship to make orbital compute possible at scale. He said each launch could carry about 10 megawatts of compute and forecast about 280 orbital compute launches by the end of 2028. If that pace is not reached by then, he said it would happen in 2029. TIKR.com adds that Musk told investors on SpaceX’s third-quarter call that he expects weekly Starship flights in the first half of 2027 and twice-weekly flights by the end of that year. At that rate, the article says, twice-weekly launches through all of 2028 would total about 100 flights for everything Starship carries, including Starlink satellites.

Spending, payback, and the value question

TIKR.com says SpaceX’s capital spending has more than quadrupled from 2023 to 2025. It also reports that SpaceX spent $48.1 billion in the first nine months of 2026 and guided to another $35 billion to $40 billion of fourth-quarter capex. On the third-quarter call, SpaceX’s CFO said new compute pays for itself in less than a year at current rental rates, but TIKR.com argues that if orbital compute costs up to 100 times as much to build while charging the same rent, the payback period would stretch from under a year to as long as a century. The article says that even at one-tenth of Google’s high-end estimate, payback would still take about a decade.

GPU makers, current financing, and what investors may value today

TIKR.com says Chanos made a separate point in response to another X user, writing that over time the profits would go to GPU suppliers, not the companies leasing the equipment or owning the real estate-like assets, which he called financial conduits. The article says SpaceX is raising $40 billion of debt that Musk described as primarily callable debt for GPU purchases, with chips supplied by NVIDIA. It adds that NVIDIA gets paid when the chips are delivered, while SpaceX depends on customers continuing to rent them. TIKR.com also notes that Anthropic agreed in its IPO filing to spend up to $120.5 billion with SpaceX through 2029, but can walk away on 90 days’ notice after its first noncancelable periods end early next year.

TIKR.com says Google is itself testing AI server chips in orbit, which means it is not a neutral observer, and it notes that Starship is intended to lower the launch costs that make space so expensive. Even so, the article says the cost gap should make investors cautious. It states that SpaceX traded at 45 times forward earnings as of Friday’s close, and says investors are already valuing Starlink and the ground-based compute SpaceX rents. The article’s view is that orbital compute should receive no value in the stock price until Starship demonstrates what a gigawatt in space will actually cost. It says the next key test is Flight 15, expected in November, when Musk expects the booster and ship to be caught, and that a launch cadence close to Maguire’s forecast would be the first clear sign the orbital model can work.