Amazon Leo Adds Uztelecom as Partner; Google's Orbital TPU Reaches Orbit

Amazon Leo, the low Earth orbit broadband network formerly known as Project Kuiper, added state-owned Uzbek operator Uztelecom to its reseller list on 2 October, according to Data Center Dynamics.

Under the deal, Uztelecom affiliate UZ-SAT becomes an authorised distributor. It will use Leo capacity to reach areas where terrestrial networks are hard to build. Uztelecom says its fiber network covers more than 378,000 km and reaches 96 percent of communities in Uzbekistan, serving more than 13 million users. The operator launched 5G in 2022 and recently added 65 base stations. "Collaborations like our agreement with Uztelecom and UZ-SAT are essential to making this a reality," Trevor Vieweg, global head of business for Amazon Leo, said in the announcement. "Our goal is to make modern digital technologies more accessible, regardless of where people live or work," UZ-SAT general director Bakhrom Azimov added.

Amazon has now launched nearly 400 of a planned 3,000 Leo satellites. It is testing the service with employees and select enterprise customers. The company expects initial commercial service in some locations later this year. Existing Leo distribution deals include Vodafone and Vodacom, Verizon, AT&T, NBN Co., Thaicom and Vrio.

The week's bigger launch

The more consequential orbital event came on Thursday 1 October. SpaceX launched Google's Project Suncatcher prototype on a Falcon 9 rideshare from Vandenberg Space Force Base in California, per CNBC. The satellite, built by Planet Labs, carries a Google tensor processing unit. It is the company's competitor to Nvidia's GPUs and the first in-orbit test of Google's plan for solar-powered AI compute in space. "Our team has confirmed contact with the satellite and it is operating as expected," Travis Beals, senior director of Project Suncatcher, wrote in a blog post cited by CNBC.

NASASpaceflight reports the Transporter-18 mission carried 130 payloads. These included NASA's SPRITE telescope and Starfish's first Otter servicing vehicle. The booster B1082-25 returned to Landing Zone 4. Google's Suncatcher M1 deployed at T+1:01:25 after the second stage completed its burn.

According to TechCrunch, the prototype will fire its TPU in 15-minute bursts to avoid straining power and thermal systems. Google is working with Planet Labs on a two-satellite demo expected next year. It will attempt a laser link between the craft. The company's longer-term design envisions 81 satellites flying in close formation.

Google's own paper says the rockets do not exist yet

The same day, Google released a peer-reviewed version of its orbital data center paper, to be published in Joule. The Register, which reviewed the document, reports that it sets a target of $200 per kilogram to orbit for space compute to be competitive with terrestrial power costs. The paper notes that SpaceX has cut launch cost per kilogram by roughly 20 percent for every doubling of cumulative mass launched. Hitting that target, Google calculates, means launching 370,000 additional tonnes of cumulative mass. That is equivalent to around 1,800 successful Starship flights with components reused 100 times.

TechCrunch puts the figure at 1,800 launches and says the paper expects launch prices near $200/kg by 2035. The two outlets describe the same arithmetic. Google's paper itself calls the outcome "plausible under reasonable assumptions," not a forecast.

"If launch costs reach less than $200/kg, annualized cost per unit of power in space could be approximately comparable to terrestrial spend."

The paper also flags unresolved problems. Existing networking gear is probably unsuitable for linking satellites into working clusters. The required formation flying is "significantly larger" and closer than any current constellation.

Capacity crunch behind the optimism

Amazon is not the only operator competing for rides. Light Reading reported on 30 September that SpaceX has reportedly stopped taking Falcon 9 orders after 2028, citing the Wall Street Journal. Elon Musk posted on X in August about winding Falcon down once Starship flies several times per week. SpaceX has not confirmed the Falcon 9 timeline.

Analysys Mason data cited by Light Reading counts more than 230 constellation projects and 1.5 million planned satellites. Most are for orbital data centers. Excluding those, 133,000 satellites are planned. At 2025's record of 5,000 satellites launched in a year, principal analyst Dallas Kasaboski said it would take almost 27 years to launch them all. Amazon Leo, AST SpaceMobile, Telesat Lightspeed and Sateliot are among the operators that need those launches.

The competition is not only Western. SpaceNews reported on 1 October that China flew its ninth and tenth launches of September. It sent the 26th batch of Guowang satellites up on a Long March 8A from Hainan on 23 September and Yaogan-40 group four on a Long March 6A from Taiyuan on 24 September. Guowang, run by state-owned China SatNet, is planned to comprise around 13,000 satellites. US Space Force tracking put the new batch in roughly circular 1,100 km orbits inclined 50 degrees. The launches took China to 70 orbital attempts in 2026. It would need at least 31 in the fourth quarter to pass 100 for the first time.

Elsewhere, Satlyt, a Sunnyvale and Nairobi software startup founded by a former Google and SpaceX product manager, told TechCrunch it raised an $8 million seed round. It flew its software on the same Transporter-18 mission. Its earlier Gemma deployment on a Momentus spacecraft cut the size of a transmission about onboard software errors by more than 60 percent, the company says.