SpaceX’s Transporter-18 rideshare mission, which launched Thursday from Vandenberg Space Force Base, carried a distinct thematic payload: testing the foundational technologies required for orbital data centers (ODCs). While rideshares typically aggregate disparate commercial and government missions, this flight focused on three critical components of space-based computing infrastructure: power generation, transmission, and processing.

What Happened

The mission deployed three specific payloads, each addressing a different segment of the ODC supply chain. Cowboy Space’s Reason-1 satellite is attempting to beam solar power to Earth using an infrared laser. This test serves as a precursor for the optical links that future Cowboy Space data centers will require. The company, which rebranded from the power-beaming startup Aetherflux in May, aims to deliver 30–100 watts to a 10–20 meter spot at a ground test facility, potentially as soon as late October. Cowboy Space describes this as the first commercial kilowatt-class laser in space designed to deliver energy to Earth.

“We’re an orbital energy infrastructure company that’s delivering power, computing power, and the ability to connect all of those things through laser infrared technology,” Cowboy Space COO and Chief Legal Officer Joseph Yaffe said.

Google’s Project Suncatcher, developed in partnership with Planet Labs, is testing the viability of running AI chips in orbit. The fridge-sized satellite carries four Trillium TPUs—roughly equivalent to one server’s worth of compute—and uses approximately 1 kW of solar power to execute Gemini queries in short bursts. Google also released an updated peer-reviewed paper in the journal Joule, detailing the project’s long-term vision for satellite constellations that could hold 50–100 kW of compute each, flying in tight, laser-linked clusters.

The third payload, Star Catcher’s Protostar, is conducting the company’s first end-to-end in-space demonstration of power beaming. It deployed a cubesat and attempted to beam measurable power to its off-the-shelf solar panels, validating commercial power-transfer operations. Star Catcher lists orbital computing companies Starcloud and Aethero among its publicized customers.

Why It Matters

The Transporter-18 mission signals a shift in the space industry from theoretical discussions about AI in orbit to practical infrastructure validation. “We passed the point of saying, ‘Will there be people doing AI computing in space?'” Yaffe told Payload. “We’re focused on building the infrastructure for that.” The success of these tests could determine the feasibility of moving energy-intensive AI workloads off Earth, potentially alleviating constraints on terrestrial data centers.

Google’s analysis suggests that economic viability depends heavily on launch costs. The company estimates that if launch costs fall to $200/kg, the annual cost of launching a kilowatt of solar power to orbit would be approximately $810, comparable to what US data centers spend on electricity. Google’s learning-curve analysis indicates this price point is possible by the mid-2030s, contingent on SpaceX’s Starship achieving a launch cadence of about 180 times per year.

Technical challenges remain significant. Google’s paper notes that while its TPUs survived a radiation dose equivalent to five years in orbit during ground tests, radiation still caused occasional silent errors—about one per three million queries. This error rate may be manageable for inference tasks but could pose problems for training models. Thermal management also remains a major hurdle; the MVP’s chips must power down every 15 to 20 minutes to avoid overheating, though Google plans for its two follow-up satellites in 2027 to run continuously.

Cowboy Space is integrating these technologies into its own rocket, whose upper stage doubles as a 1 MW data center, targeted for launch in December 2028. The same laser technology tested on Reason-1 will form the backbone of the data link on Reason-2, which is slated to fly H200 GPUs in the first half of 2027, beaming data rather than power.

The Bottom Line

Transporter-18 did not deliver a fully operational orbital data center, but it successfully deployed three distinct experiments testing the power, compute, and transmission pillars of the concept. As Yaffe noted, the “if” of AI compute in space may be settled; the coming weeks will provide initial data on how far along the “how” is.