Weekly Brief.

Axiom’s October 6 update adds an operational checkpoint for its two Axiom Resilient Compute nodes, launched earlier this year on Kepler’s optical-relay satellites. The company says the nodes are operational and confirms communications between them and with the ground.

The cryptography result in the same announcement needs a separate reading. Axiom says it migrated an encrypted workload using post-quantum cryptography in a ground demonstration; the orbital demonstration is still ahead. The release gives no measured compute throughput, paid utilization or date for that next test. The release does not yet quantify commercial compute usage.

Unibap and Leonardo said on October 5 that they will develop an adapted iX20 computing platform for Leonardo’s Space Guardian Earth-observation constellation. The intended applications include defence and emergency response, with processing closer to the observations.

Their relationship dates back to a joint-study announcement in October 2025. This week’s release supplies a clearer hardware and mission scope, but leaves order volume, price and delivery timing undisclosed. It belongs in the development pipeline; readers should wait for procurement and flight milestones before treating it as delivered capacity.

In an October 8 account of its October 5 IAC presentation, STAR.VISION described the Marco Polo and Ibn Battuta compute pair and two additional satellites developed with Oman Lens. Its architecture assigns complementary decision-making and computing roles, connected through optical links.

That is a useful statement of design direction. The release provides no individual launch dates, commissioning records or independently measured performance for the four presented satellites. It cannot support adding four newly operating nodes to a market total.

This week’s register additions include Privateer Pono and ReOrbit Space Cloud. Privateer’s December 2023 account records a hosted compute-and-communications prototype aboard D-Orbit’s ION. ReOrbit’s June 2026 announcement reports €4.6 million from Business Finland for OrbitCloud research combining connectivity and orbital AI inference.

Both improve the historical record. Neither is an October launch or funding round. ReOrbit’s projected programme value above €40 million should also remain separate from the €4.6 million in funding it announced.

Connected nodes and mission-specific processing give orbital computing clearer near-term jobs. Axiom’s communications result, Unibap’s development scope and STAR.VISION’s architecture announcement sit at different stages of that progression. Combining them into one capacity figure would erase the distinctions that make the evidence useful.

The commercial questions remain specific: which workloads run reliably, what customers commit to buy, and whether processing in orbit saves enough time or communications cost to justify the hardware.

Watch for Axiom’s orbital cryptography test, Unibap–Leonardo delivery commitments and mission-specific evidence behind STAR.VISION’s claims. Reported workloads, elapsed processing time and customer use would make the next round of announcements more informative. No unconfirmed item is included in this edition.

Google said its Project Suncatcher prototype launched on October 1 aboard SpaceX’s Transporter-18 mission. The company confirmed contact and said the Planet-built satellite was operating as expected. It will now gather data on how Google’s TPUs handle launch stress, radiation and thermal extremes.

That is real progress, but carefully bounded. Google has not yet reported a completed in-orbit AI workload, sustained thermal performance or a satellite-to-satellite link. This mission tests components; it does not establish a commercial computing service.

Punjab Newsline reported on October 2 that TakeMe2Space’s MOI-1A launched on the same Transporter-18 rideshare. Before launch, Reuters described a sub-50 kg spacecraft carrying Nvidia Orin NX processors and using about 150 watts. TakeMe2Space told Reuters it had signed 23 customers; those are company-reported bookings, not verified usage or revenue.

MOI-1A is better understood as edge computing aboard a satellite than as a terrestrial data center in miniature. At the cutoff, TakeMe2Space’s public mission page did not yet show deployment, contact or commissioning evidence. The meaningful next proof is a customer workload completed and returned from orbit.

Firefly announced on September 30 that Starcloud will fly its SC-1L processor on an Elytra vehicle during Blue Ghost Mission 3, targeted no earlier than 2028. After the NASA lunar-surface phase, the payload is intended to process imagery from Firefly’s Solux system in lunar orbit.

This is a commercial hosting agreement, not merely supplier interest. It still leaves price, processor configuration, computing capacity, launch vehicle and exact date undisclosed. The deal adds a credible route to a lunar demonstration; it does not add operating capacity today.

SignalSpore now tracks 54 projects and 226 published supplier records, up from 23 and 52 in the opening snapshot. The 31 added profiles include operating or flown edge-compute systems as well as proposals; most of the underlying events predate this week. The supplier expansion is similarly dominated by historical backfill.

Twenty records are now classified as awarded contracts. The other 206 include supplier selections, partnerships, development agreements, preliminary agreements and one capacity reservation. The larger register improves market visibility, but it should not be mistaken for 174 new orders.

This week’s strongest signals are not interchangeable. Suncatcher is testing whether AI hardware survives and runs in orbit. MOI-1A is aimed at practical, low-power data processing close to the sensor. SC-1L is a booked, longer-range lunar demonstration. Calling all three “space data centers” hides more than it explains.

The common weakness is commercial evidence: little disclosed pricing, sparse utilization data and few measured workloads. Hardware is getting easier to point to. Repeatable demand and operating economics are not.

Watch for Suncatcher TPU power-on and workload results, MOI-1A deployment and customer-use evidence, and firmer SC-1L mission details. We reviewed the editorial backlog and found no current, attributable unconfirmed claim strong enough to include this week.

Google’s September 24 update names SpaceX’s Transporter-18 mission for its first orbital TPU experiment, developed with Planet. The immediate job is to test hardware survival and cooling. A two-satellite networking test is planned for 2027.

That gives suppliers and observers a concrete milestone. It does not establish a commercial service, a disclosed contract value, or the economics of running a large computing fleet.

Beijing’s municipal portal reports that Chaozhisuan-1 launched on September 20. The operator describes onboard imagery processing and AI inference, with ambitions for a wider constellation. The reviewed announcement does not disclose a benchmark, hardware suppliers or a funded fleet size.

The distinction matters: launch is a milestone we can record; sustained computing performance is the next thing to look for.

The opening register now covers 23 projects. Recent additions include China’s Chaozhisuan-1 and Chenguang-1, Japan’s SpaceBlast demonstration, Altair-Next Gen and Blue Origin’s proposed Project Sunrise. Several rely on announcements predating this week; adding them closes research gaps rather than signalling new launches.

Fourteen projects have linked supplier records; nine do not. Across the full supplier register, partnerships account for 30 of 52 records. The balance is six contract awards, eleven supplier selections, three development agreements, one capacity reservation and one preliminary agreement. Big programme ambitions should not be read as firm supplier backlog.

DIGITIMES reported on September 24 that Japan’s EdgeCortix is seeking a SpaceX collaboration on chips for space data centers. That is a reported ambition, not evidence that SpaceX has agreed, placed an order or selected a chip.

It is worth watching because it puts a name to the competition for future orbital-computing work. A statement from both companies, a defined scope or a procurement announcement would give this lead more weight. For now, it stays outside our supplier totals.

Updated 25 September: this item replaces the earlier rumor note after reviewing additional reporting available before the cutoff.

The near-term story is increasingly about hardware, integration and experiments. The leap from those steps to a competitive computing service remains large. Our next markers are Suncatcher’s flight and test results, published Chaozhisuan-1 performance, and supplier commitments that turn proposals into work.

The most useful evidence may be a modest repeat order or a measured workload. Neither needs a giant constellation announcement to matter.