# Will SpaceX launch at least one of its AI1 data-centre satellites before 1 January 2028?

> As of 17 November 2026, Prolepsia puts the chance of yes at 60%.

- Forecaster: Prolepsia (https://prolepsia.com)
- Forecast made: 2026-11-17
- Question type: Yes / no
- Resolves: 2028-01-01
- Status: Open
- Forecasts: 1
- Page: https://prolepsia.com/forecasts/stratus-phase-momentum/will-spacex-launch-at-least-one-of-its-ai1-data-centre-satellites-before-1-janua
- Open in the app: https://app.prolepsia.com/library/3773c7f9-9ceb-499f-92d6-1d88b366a23c

## Forecast

- Yes: 60%
- No: 40%

## How Prolepsia reasons

## TL;DR
Probability of Yes: 60%. Probability of No: 40%. The deciding milestone is getting one identifiable AI1 demonstrator through integration and safely separated in space before January 1, 2028, not getting the satellite to work or starting a fleet rollout.

## Context
SpaceX is developing AI1, also known as Starmind, to run substantial AI workloads in space. On January 30, 2026, it applied to the FCC for permission to launch and operate up to one million orbital data-centre satellites. The [FCC accepted the application for filing on February 4](https://docs.fcc.gov/public/attachments/DA-26-113A1.pdf). That was not an authorization. As of November 17, 2026, no AI1 launch had been publicly established.

This forecast concerns one spacecraft, not the full system. An identifiable AI1 prototype or direct successor carrying meaningful AI computing hardware counts if it separates above 100 km before 00:00 UTC on January 1, 2028. Any launcher can qualify. The satellite need not reach its intended orbit or function after separation. January 31, 2028 is the cutoff for confirmation, not an extension of the flight window.

## Evidence
Prototype history shows both how quickly a small test can fly and how easily a target can slip. Google’s [Project Suncatcher](https://blog.google/innovation-and-ai/technology/research/google-project-suncatcher?trk=public_post_comment-text) put one compute demonstrator on Falcon 9 on October 1, 2026, 331 days after its public announcement and just seven days after Google publicly named the mission. Its planned two-satellite test was a separate milestone. [Starcloud](https://blogs.nvidia.com/blog/starcloud?trk=public_post_comment-text) flew a much smaller 60 kg satellite carrying one H100 in November 2025. Those flights show that a single compute prototype can precede a more ambitious system. They do not show that a rack-scale AI1 is equally easy to finish.

The late cases matter because the deadline falls at the end of SpaceX’s target year. [Amazon’s Kuiper prototypes](https://www.aboutamazon.com/news/innovation-at-amazon/project-kuiper-announces-plans-and-launch-provider-for-prototype-satellites?linkCode=w50&ots=1&tag=arstech20-20) moved from a fourth-quarter 2022 target to an October 6, 2023 launch and changed launch vehicles. [AST SpaceMobile’s BlueWalker 3](https://ast-science.com/wp-content/uploads/2023/03/asts-20221231.htm.pdf) moved from late in the fourth quarter of 2021 to March 2022 and then flew in September 2022. SpaceX itself [told Congress in May 2017](https://www.commerce.senate.gov/wp-content/uploads/media/doc/Testimony%20Patricia%20Cooper%20-%20Broadband%20Infrastructure%20Hearing.pdf?utm_source=openai) that it planned one Starlink prototype by year-end and a second in early 2018. Both Tintin satellites flew on February 22, 2018, according to the [FCC’s constellation order](https://docs.fcc.gov/public/attachments/FCC-18-38A1.pdf). These are comparisons, not an on-time launch rate. They show why even a short slip is material for AI1.

SpaceX’s own language shifted over 2026. In a [June 8 interview filed with the SEC](https://www.sec.gov/Archives/edgar/data/1181412/000162828026041761/fwp_cfointerview.htm), CFO Bret Johnsen distinguished flying pieces in 2027 from flying satellites in 2028. The [June 12 prospectus](https://www.sec.gov/Archives/edgar/data/1181412/000162828026042639/spaceexplorationtechnologi.htm) said deployment could start as early as 2028, and [May SEC correspondence](https://www.sec.gov/Archives/edgar/data/1181412/000162828026036964/filename1.htm) shows that the 2028 wording was deliberate. Musk gave a fourth-quarter 2027 target in [August](https://ticker-brief.com/stocks/spcx/2026-08-24-posts). In [September conference remarks](https://docs.publicnow.com/viewDoc?filename=28442%5CEXT%5CD2382E66D06B2CAD0252E23426FEE3F569C3482C_7A6291CA6B33DCE0B66186B170E415BE35AF695C.PDF), Johnsen explicitly targeted orbital-compute satellites in 2027 and described reusing much of the Starlink V3 platform. On [October 20](https://s21.q4cdn.com/184289198/files/content_files/SpaceX-Third-Quarter-2026-Webcast-Transcript.pdf), Gwynne Shotwell named Starmind and said flights would begin in 2027. Musk offered the third quarter on November 6, but called it a guess. The November 17 TomorrowX recap repeated the 2027 target without reporting a new hardware milestone. The later remarks speak more directly to first flight than the June prospectus does. They express a shared company plan, not independent evidence that the schedule is firm.

Hardware disclosures point to real work, not a ready spacecraft. [NVIDIA confirmed in August](https://investor.nvidia.com/news/press-release-details/2026/SpaceXAI-Adopts-NVIDIA-Vera-CPU-to-Accelerate-Agentic-AI-at-Massive-Scale/default.aspx) that first-generation Starmind was planned around an adapted Vera Rubin NVL72 system. The [June design](https://www.datacenterdynamics.com/en/news/spacex-details-ai1-satellite-data-center-claims-150kw-peak-compute) described 120 kW average and 150 kW peak compute power; a later [SpaceX programme page](https://www.spacex.com/spacexai/starmind) listed 175 kW average and up to 250 kW peak. Neither set of figures is a flight result or a threshold for YES. On November 10, [Singulus announced an order worth more than €60 million](https://econotimes.com/Singulus-Secures-SpaceX-Solar-Equipment-Order-Worth-More-Than-60-Million-1775622) for its first solar-cell equipment line, due for delivery in the first half of 2027. That is an order for factory equipment, not flight-qualified arrays. The evidence does not establish that one prototype must wait for the line.

Starship has cleared a separate hurdle: releasing real satellites. Flights 10 and 11 released eight simulators each. [Flight 12](https://www.spacex.com/launches/starship-flight-12) carried 20 simulators and two real camera satellites. [Flight 13](https://www.spacex.com/launches/starship-flight-13/) released 20 real Starlink V3 satellites on a trajectory that reentered. Flights 14 and 15 deployed 26 V3 satellites each into orbit on September 28 and November 13, respectively ([SpaceX quarterly results](https://s21.q4cdn.com/184289198/files/content_files/SpaceX-Reports-Third-Quarter-2026-Results.pdf?aff_unique2=c10coZOPj&code=c10coZOPj); [Flight 15 report](https://www.spacex.com/launches/starship-flight-15)). Those are two orbital deployment missions, not 52 independent tests. Flight 15’s booster recovery failed, but payload separation succeeded. The record supports Starship as a route for one flight. It does not demonstrate the AI1 separation interface or establish that a smaller AI1 is compatible with Falcon.

## What's non-obvious
The June prospectus is neither a veto on a 2027 flight nor something to discard. A dedicated demonstrator can precede deployment, which is why Shotwell’s spacecraft-specific wording matters. But the June distinction between pieces and satellites still points to a concrete way for SpaceX to make progress in 2027 without resolving YES. A Starlink carrying experimental compute hardware does not count on that basis alone. It needs explicit identification as an AI1 programme demonstration. Conversely, a smaller dedicated AI1 can count even if it fails immediately after release.

Fleet authorization is also not necessarily the gate for a single flight. SpaceX received [experimental authority for Tintin in November 2017](https://fcc.report/ELS/Space-Exploration-Technologies-Corp/0298-EX-CN-2016), and the prototypes flew before the [March 2018 constellation grant](https://docs.fcc.gov/public/attachments/FCC-18-38A1.pdf). That precedent makes a narrow test authorization plausible. It does not show that SpaceX has obtained one for AI1.

## Uncertainties
The main gap is flight readiness. As of November 17, there was no verified integrated AI1 flight article, environmental or separation qualification result, or assigned mission. Searches for AI1 and Starmind in public launch listings and [satellite catalogues](https://celestrak.org/satcat/records.php?NAME=STARMIND&FORMAT=json) did not turn up named matches. That does not rule out an internal booking, a different payload name, or a late announcement. The TomorrowX remarks were not verified against the organizer’s recording, so the recap adds little to the October signal.

The other gap is mission-specific clearance. The [FCC filing notice](https://docs.fcc.gov/public/attachments/DA-26-113A1.pdf) establishes acceptance, not a grant. No separate AI1 authorization was verified, and no Falcon assignment was confirmed. The strongest positive update would be a named flight article completing integrated tests, paired with an authorization or launch window. If there is still no sign of integrated hardware late in the third quarter of 2027, there will be less room to finish qualification, arrange a flight, or recover from a failed attempt. That is the central downside: a delay of a few months can turn a real programme with a credible first flight into a NO on this deadline.

## Forecast history

| Date | Forecast |
|---|---|
| 2026-11-17 | 60% chance of yes |

## Sources Prolepsia read

- [# PUBLIC NOTICE](https://docs.fcc.gov/public/attachments/DA-26-113A1.pdf)
- [Project Suncatcher explores powering AI in space](https://blog.google/innovation-and-ai/technology/research/google-project-suncatcher?trk=public_post_comment-text)
- [Starcloud](https://blogs.nvidia.com/blog/starcloud?trk=public_post_comment-text)
- [Amazon’s Kuiper prototypes](https://www.aboutamazon.com/news/innovation-at-amazon/project-kuiper-announces-plans-and-launch-provider-for-prototype-satellites?linkCode=w50&ots=1&tag=arstech20-20)
- [AST SpaceMobile’s BlueWalker 3](https://ast-science.com/wp-content/uploads/2023/03/asts-20221231.htm.pdf)
- [Microsoft Word - SpaceX Testimony Senate Commerce Committee Broadband Rev.05.01.2017](https://www.commerce.senate.gov/wp-content/uploads/media/doc/Testimony%20Patricia%20Cooper%20-%20Broadband%20Infrastructure%20Hearing.pdf)
- [June 8 interview filed with the SEC](https://www.sec.gov/Archives/edgar/data/1181412/000162828026041761/fwp_cfointerview.htm)
- [August](https://ticker-brief.com/stocks/spcx/2026-08-24-posts)
- [Edited Transcript](https://docs.publicnow.com/viewDoc?filename=28442%5CEXT%5CD2382E66D06B2CAD0252E23426FEE3F569C3482C_7A6291CA6B33DCE0B66186B170E415BE35AF695C.PDF)
- [October 20](https://s21.q4cdn.com/184289198/files/content_files/SpaceX-Third-Quarter-2026-Webcast-Transcript.pdf)

## The question

This question asks whether SpaceX will successfully launch at least one satellite belonging to its first-generation AI1 orbital data-centre programme before 1 January 2028. The forecasting period begins on 17 November 2026 and ends on 31 December 2027, inclusive.

SpaceX is developing a constellation of satellites designed to provide artificial intelligence computing capacity in orbit, potentially reducing reliance on terrestrial data centres. On 30 January 2026, SpaceX submitted an application to the US Federal Communications Commission (FCC) seeking permission to launch and operate up to one million satellites as part of its Orbital Data Center system. The FCC accepted the application for filing on 4 February 2026. The proposed system would operate primarily in low Earth orbit, using solar power and optical inter-satellite communications. Source: FCC Public Notice DA 26-113, February 2026. ([docs.fcc.gov](https://docs.fcc.gov/public/attachments/DA-26-113A1.pdf))

In June 2026, SpaceX revealed details of its first-generation AI1 satellite concept, subsequently associated with its Starmind orbital computing programme. The AI1 design incorporates high-performance AI computing hardware, substantial solar power generation and dedicated thermal management systems. SpaceX has described a computing payload of approximately 120 kW, with peak capacity reaching 150 kW. The satellites are intended to perform significant AI computing workloads in space rather than simply relay communications. ([datacenterdynamics.com](https://www.datacenterdynamics.com/en/news/spacex-details-ai1-satellite-data-center-claims-150kw-peak-compute/?trk=article-ssr-frontend-pulse_little-text-block))

SpaceX previously indicated that its first AI1 satellites could launch during 2027. However, its June 2026 prospectus stated that deployment of orbital AI computing satellites was expected to begin as early as 2028, highlighting uncertainty around the development and deployment schedule. As of 17 November 2026, publicly available information does not establish that an AI1 satellite has launched. ([datacenterdynamics.com](https://www.datacenterdynamics.com/en/news/spacex-ipo-musks-firm-set-to-launch-first-orbital-data-center-ai1-satellites-in-2027-will-put-compute-on-starlink-craft/))

The question specifically concerns whether SpaceX launches at least one physical AI1 satellite before the deadline. It does not require the deployment of a complete constellation, the commencement of commercial operations or the successful provision of AI computing services.

### How it resolves

The question resolves YES if at least one satellite identifiable as part of SpaceX's AI1 first-generation orbital data-centre programme is launched from Earth before 00:00 UTC on 1 January 2028.

For a YES resolution, the satellite must be an actual spacecraft developed under the AI1 programme, including a flight prototype, demonstration unit or production satellite. It must carry hardware intended to provide substantial AI computing capabilities in orbit, rather than merely conventional communications equipment. The launch must achieve successful separation of the satellite from its launch vehicle in space, whether or not the satellite subsequently reaches its intended operational orbit.

The satellite does not need to demonstrate functioning AI computing hardware, establish communications with the ground, generate its intended electrical power or begin commercial operations. A satellite that is successfully deployed but subsequently malfunctions or becomes inoperable still qualifies.

The satellite may be launched aboard SpaceX's Starship, Falcon 9, Falcon Heavy or another launch vehicle. The identity of the launch provider does not affect resolution, provided the payload is genuinely a SpaceX AI1 satellite. Launches from any location worldwide qualify.

The question resolves NO if no qualifying AI1 satellite is successfully launched and separated from its launch vehicle before the deadline. This includes circumstances in which SpaceX announces, manufactures, tests or schedules an AI1 satellite but does not actually launch and deploy it before 1 January 2028. A launch attempt that fails before satellite separation does not qualify.

Resolution will primarily rely on official SpaceX announcements, mission descriptions, launch reports and company disclosures identifying the payload and confirming its launch. Independent confirmation may be obtained from established spaceflight tracking and reporting organisations, including Jonathan McDowell's launch records, CelesTrak, the US Space Force satellite catalogue and reputable aerospace news publications such as SpaceNews or Ars Technica.

A launch can qualify even if SpaceX does not publicly announce it immediately, provided credible evidence subsequently establishes that an AI1 satellite was successfully launched and separated before the deadline. The outcome should be determined using evidence available by 31 January 2028, allowing time for delayed announcements and confirmation of launches occurring near the deadline.

If credible evidence establishes that a qualifying launch occurred, the question resolves YES even if some details of the satellite's technical capabilities remain undisclosed. If no qualifying launch can be established by the verification cutoff, the question resolves NO. If credible sources establish that an unidentified payload was launched but leave unresolved whether it was an AI1 satellite, and this uncertainty could change the outcome, the question will be annulled rather than resolved based on speculation.

### Fine print

The term 'AI1' refers specifically to SpaceX's first-generation orbital AI data-centre satellite design, also associated with the Starmind programme. A satellite does not have to retain the exact AI1 branding if SpaceX renames the programme before launch, provided reliable evidence establishes that it is the same first-generation satellite programme or its direct successor.

Ordinary Starlink communications satellites do not qualify merely because they contain onboard processors, AI-assisted networking capabilities or other computing equipment. Similarly, a Starlink satellite equipped with additional experimental AI hardware does not automatically qualify unless SpaceX explicitly identifies it as an AI1 spacecraft or as a flight demonstration of the AI1 satellite programme.

A dedicated AI1 prototype or demonstration satellite qualifies even if it carries less computing capacity than the originally announced design, provided it is clearly identified as a flight article within the AI1 programme and carries meaningful AI computing hardware.

A satellite developed by another company, including a customer payload launched aboard a SpaceX rocket, does not qualify solely because SpaceX provides launch services. The payload must belong to SpaceX's own AI1 programme.

Ground-based demonstrations, structural test articles, suborbital tests that do not deploy a satellite into space, and launches carrying only mass simulators do not qualify. A qualifying spacecraft must separate from its launch vehicle above the internationally recognised 100 km Kármán line. It need not remain in orbit afterward.

All deadlines use Coordinated Universal Time (UTC). A launch occurring on 31 December 2027 UTC qualifies, whereas one occurring on or after 1 January 2028 UTC does not. For this question, the relevant event is successful spacecraft separation, rather than rocket liftoff. Thus, a rocket lifting off before the deadline but deploying its satellite afterward does not qualify.

---

Prolepsia’s forecasting engine made this forecast: it researched the question, weighed the evidence and wrote the report on this page. A probability is not a promise; this one changes as the evidence does.
