SpaceX (SPCX.US) CEO Elon Musk announced last month that the first batch of AI computing satellites equipped with Nvidia (NVDA.US) chips will launch into orbit in the fourth quarter of 2027, moving the timeline up yet again from the “earliest 2028” date stated in the company’s IPO prospectus. However, multiple satellite industry experts and investors believe that orbital data centers will need at least five to seven more years to reach true commercial scale, with the 2030s emerging as a more pragmatic timeline.

Musk posted on X in late August that SpaceX has collaborated with Nvidia to complete the design of the Vera Rubin NVL72 system optimized for the space environment, with plans to send it into orbit in Q4 2027 and achieve “significant scale” deployment in 2028. He described the satellite data center as “simpler, lower cost, higher density, and lighter weight than traditional racks.”

The same day, Nvidia officially confirmed that SpaceX AI will deploy next-generation Vera CPUs to accelerate agentic AI workloads, and will send an optimized version of the Vera Rubin NVL72 rack-scale system into orbit as the computing core of the first-generation Starmind satellites. This collaboration marks a critical inflection point as the world’s first space-based AI data center moves from concept to engineering implementation.

This is already the second time SpaceX has accelerated its timeline this year. The company targeted “earliest 2028” in its May IPO prospectus, Musk said “next year” during the August 4 earnings call, and now the timeline has been further narrowed to a specific Q4 2027 date.

Industry Consensus Points to the 2030s

Evelyn Chow, portfolio manager at Neuberger Berman, said bluntly: “This is something that will happen in the next decade — the 2030s. Over the next four to five years, we need to see a massive number of satellite launches, along with the connectivity infrastructure built in parallel, before we can even begin to think about orbital data centers truly achieving scale.”

Blaine Curcio, founder of Hong Kong-based satellite market research firm Orbital Gateway Consulting, also agreed that the 2030s is a “reasonable assessment.” However, he cautioned that SpaceX has repeatedly proven investor skepticism wrong. “If you had asked any satellite industry expert in the late 2010s whether SpaceX could launch 10,000 satellites by 2025, including myself, every single one would have said ‘absolutely impossible.'”

Four Major Technical Bottlenecks
Heat Dissipation and Radiation Resistance

Chow pointed out that ground-based data centers can rely on liquid cooling technology to manage heat, but in the vacuum of space, heat transfer mechanisms are fundamentally different. She emphasized: “Heat dissipation is a massive problem.” Additionally, orbital hardware must possess extreme radiation resistance to operate long-term in the harsh space environment.

GPU Iteration Speed

Curcio believes the pace of technological advancement itself poses a barrier. “GPUs are evolving extremely fast. If you launch a state-of-the-art data center into space at enormous cost, it could become obsolete within just a few years.”

Massive Data Downlink

Transcelestial, a Singapore-based company specializing in laser communication technology, is currently working to solve another core challenge: how to reliably transmit massive amounts of data back to Earth. Co-founder and CEO Rohit Jha said: “Anyone can build a data center, but if you cannot communicate efficiently with these AI systems, those data centers are useless.”

Hyperscale Power Supply

Jha added that orbital data centers may need another five to seven years to reach “hyperscale” status, and supporting that level of computing power will very likely require nuclear power generation in the future.

Business Model and Financial Pressure

The cost ledger for orbital data centers is under intense scrutiny from Wall Street. Commercial consultancy Wood Mackenzie estimates that building a 1-gigawatt (GW) orbital data center would cost approximately $170 billion (about NT$5.4 trillion), more than triple the cost of a traditional ground-based data center. Morgan Stanley analyst Adam Jonas projects that SpaceX’s AI-related capital expenditures will reach $53 billion (about NT$1.7 trillion) in 2026, with an additional $130 billion (about NT$4.1 trillion) in 2027.

Amazon founder Jeff Bezos and researcher Andrew McCalip, among others, have pointed out that expensive AI chips and high launch costs constitute current industry barriers, and the economic model at this stage is not yet rational.

However, SpaceX is not entirely without demand-side support. According to CNBC, SpaceX President Gwynne Shotwell revealed before the IPO that the company has signed orbital computing power lease agreements with AI startup Anthropic and Google. This means that even before the first satellite has launched, major AI customers have already reserved computing capacity.

MetricFigureFirst Starmind satellite launchQ4 2027Large-scale deployment target2028Industry consensus for scale2030sOrbital data center build cost (1GW)approximately $170 billion2026 AI capex estimate$53 billion2027 AI capex estimate$130 billion

Note: Capital expenditure figures are Morgan Stanley analyst estimates; build cost is a Wood Mackenzie estimate.

JPMorgan analyst Doug Anmuth offered a more optimistic long-term outlook, projecting that SpaceX could deploy approximately 75 GW of orbital computing power by the end of 2031, with costs significantly lower than ground-based construction, making it a major new revenue source beyond the company’s launch and Starlink businesses.

Musk’s Track Record on Timelines

Musk’s history with timelines is the primary reason investors remain skeptical about the specific “Q4 2027” date. The Tesla Roadster, unveiled in 2017, has still not shipped, and both the Cybertruck and Tesla Semi arrived years behind their original targets. This track record has led the market to treat any specific quarter Musk proposes as “directional guidance” rather than a “commitment.”

Furthermore, multiple execution items remain outstanding. The program requires an entirely new satellite architecture, successful orbital deployment, and regulatory approval for a constellation of up to 1 million satellites. SpaceX applied to the U.S. Federal Communications Commission (FCC) earlier this year for a license to launch 1 million data center satellites, with deployment spanning low Earth orbit at altitudes of 500 to 2,000 kilometers. Any failure in the chain could delay the first launch or prevent SpaceX from reaching the scale needed to make orbital computing a core business.

How the Market Views SpaceX’s Big Bet

According to the Insider Monkey database, SpaceX was held by 119 hedge funds in the second quarter of 2026 (the first quarter with hedge fund holdings data following the company’s June IPO), with total holdings valued at $116.45 billion (about NT$3.7 trillion), making it one of the largest dollar positions in the entire database. There is no directly comparable pure-play launch or orbital computing competitor in the same dataset, reflecting SpaceX’s unique position across the industries it spans.

Supporters argue that space-based data centers can circumvent the regulatory hurdles and community opposition facing ground-based facilities, while SpaceX is attempting to break down cost barriers through vertical supply chain integration. On one hand, the company relies on the Starship heavy-lift rocket to dramatically reduce launch costs; on the other, it is developing proprietary AI chips through a chip fabrication facility called Terafab, in collaboration with partners Tesla and Intel. By controlling both launch and underlying computing hardware, SpaceX is accelerating the commercialization of space-based computing.

However, unit economics remain undisclosed. Musk claims orbital rack costs will be lower than traditional ground-based infrastructure, but SpaceX has yet to reveal sufficient detail to prove whether these savings will offset the costs of launch, power, cooling, maintenance, and replacement of orbital computing hardware. Without verifiable unit economics data, investors cannot determine whether orbital data centers will ultimately become a high-margin business or an expensive technology experiment.