Google and Blue Origin sign launch agreement for TPU satellites in sun-synchronous orbit

Google and Blue Origin have signed an agreement for the New Glenn rocket to carry Project Suncatcher satellites carrying Google TPUs into orbit in the first quarter of 2029 — the project's first dedicated heavy-lift launch, two years after…

File photo: File photo: Giant Android robot at the Googleplex, Mountain View, California.
File photo: secretlondon123, CC BY-SA 2.0 · source This adaptation is licensed under CC BY-SA 2.0.

Google and Blue Origin sign launch agreement for TPU satellites in sun-synchronous orbit

Google and Blue Origin have signed an agreement for the New Glenn rocket to carry Project Suncatcher satellites carrying Google TPUs into orbit in the first quarter of 2029 — the project's first dedicated heavy-lift launch, two years after the prototype was launched on a Falcon 9.

Google and Blue Origin announced on Thursday, 19 November 2026, that they have entered into a launch services agreement: Blue Origin's New Glenn rocket will carry satellites equipped with Google's Tensor Processing Units (TPUs) into orbit in the first quarter of 2029 (The Verge). The mission concerns Project Suncatcher, Google's effort to place AI computing in space. Google itself announced the partnership in a blog post about the project's "first heavy-lift launch" (letsdatascience).

What the Agreement Involves

By the companies' own description, this will be Project Suncatcher's "first dedicated heavy-lift launch" — in contrast to the prototype, which flew as part of a shared launch. The satellites are to be placed in sun-synchronous orbit, and the companies say each satellite will have "computing capacity at roughly the scale of a data center rack" (The Verge). That is a company claim, not an independently verified specification.

Jordan Charles, senior vice president for New Glenn at Blue Origin, says the rocket's "heavy-lift capacity and large payload volume give customers the flexibility to bring new capabilities to orbit" (Unite.AI).

How many satellites the launch involves, what the contract is worth, and at what orbital altitude they will be placed is not stated in any of the available source material.

What the Project Has Achieved So Far

The milestone line begins in October 2026. A prototype satellite reached orbit on 1 October 2026 after being launched on a SpaceX Falcon 9 as part of a rideshare launch alongside several other payloads (The Verge). The spacecraft carried four TPUs, and its task was to run limited AI inference and to test how the hardware withstands the harsh conditions of space (Cryptobriefing).

As of 30 October 2026, Google reports that the host and TPU trays are powered up and that early machine learning tests are underway during commissioning. The onboard Gemma 4 model generated a haiku that was sent down to Earth — something Google describes as the first AI response from a data center-class TPU in orbit (Unite.AI). The haiku detail is in practice a marker that inference worked aboard; the actual tests were limited.

The Road to 2029

According to Google's stated plan, the next steps consist of thermal stress and radiation testing of the hardware, followed by a planned milestone in 2027: two interconnected satellites in orbit (Unite.AI). Interconnection in orbit is a necessary step toward anything resembling a distributed computing constellation — in a data center, racks are linked with high bandwidth, and equivalent linking between satellites is one of the technical problems not yet solved.

The New Glenn launch in the first quarter of 2029 is the next major milestone after that. But the project is still in a research phase, and Google has not announced any timeline for a fully operational constellation beyond the 2029 launch (Cryptobriefing).

The Energy Rationale — a Company Claim

The entire rationale for placing AI hardware in orbit rests on an energy calculation presented by Google itself: satellites in low orbit have near-continuous access to sunlight and, according to Google, can generate up to eight times more solar power than an equivalent installation on Earth (Unite.AI). This is the project's energy rationale, but it is a company calculation that has not yet been independently confirmed — and it says nothing about the cost of getting the hardware up there or the radiation and thermal wear required to keep it operational.

What Remains Unresolved

Several central questions lack answers in the public material:

  • Scale. The number of satellites in the 2029 launch, the contract value, and the orbital altitude have not been disclosed. The rack-scale claim per satellite is the companies' own characterization.
  • The timeline. A launch window more than two years out is aspirational. New Glenn has no long documented launch cadence yet, and no contract penalties or backup arrangements are known. That this is the project's "first dedicated heavy-lift launch" makes the dependence on one rocket and one provider clear.
  • Operational phase. There is no announced timeline for an actual operational constellation. Between the 2029 launch and a scenario in which computing is genuinely delivered from orbit lie thermal and radiation testing, satellite interconnection, and unresolved questions about maintenance and lifespan.

The companies present the project as a step toward space-based AI computing. How it compares to Earth-based data centers in cost, reliability, and computing actually delivered is not documented by anything in the available source material — that is precisely what the next in-orbit tests are meant to provide data on.

AIMag.no
AIMag.no
The AIMag.no editorial team covers artificial intelligence, tools, research, and regulation.

Get the best of AI MAG in your inbox

News, analysis, and ideas at the intersection of AI and society.