Welcome to Issue 8 — read why FCC was asked to pause the orbital data-center rush
The market for AI-driven Earth observation consolidated this week through buyouts, new supply, and government demand, while the orbital-data-center thesis drew its first regulatory pushback.
On the AI4Space side, the Earth-observation value chain moved on every layer. MDA Space made a firm €567 million offer for a roughly 70% stake in French analytics firm CLS, and EarthDaily brought its constellation to the commercial-operations threshold with the latest launch. Autonomy closed two loops — CAPSTONE ended nearly four years of cislunar optical-navigation work, and JPL's ERNEST rover completed a 26-kilometer autonomous desert traverse — while government demand followed from both the NGA's change-detection solicitation and the UK Space Agency's EO innovation call.
On the Space4AI side, the orbital data-center race drew profile and pushback in the same week. CNN profiled Starcloud at a $1.1 billion valuation ahead of its Blackwell satellite, while a coalition of environmental and scientific groups asked the FCC to pause processing orbital data-center applications amid a broader push for megaconstellation oversight.
On both sides, the story shifted from ambition to institutional response, regulators being challenged on orbital compute, buyers and agencies moving on EO AI.
Specifics below.
Did you know?
We have recently launched the Bits & Orbits Insights
Check out the growing set of independent dashboards, trackers, and maps on the
AI × Space frontier:
AI4Space
EarthDaily launches EDC-08 on Transporter-17, reaching commercial operations threshold for AI-ready EO constellation
On July 9, EarthDaily announced successful initial contact with EDC-08, which launched aboard SpaceX's Transporter-17 mission on July 7. The Vancouver-based company said the event, its third launch ("Launch III") for the EarthDaily Constellation, brings the system to the satellite count required for commercial operations.
Like each spacecraft in the constellation, EDC-08 carries 16 imagers collecting data across 22 spectral bands. Eight satellites are now in orbit, with seven progressing through commissioning after the May deployment of EDC-02 through EDC-07, which are returning initial imagery released alongside the announcement.
The constellation is built for broad-area change detection using AI and machine learning, optimized for time-series analysis and direct integration into automated analytics that flag change across wide areas, from aircraft and ships in harbors to roads and infrastructure. CEO Don Osborne framed the design around data consistency: "Variability in atmospheric conditions, viewing geometry, and sensor calibration can introduce noise that compounds across time-series analysis and makes data harder to trust in AI workflows." He said EarthDaily addresses this through tightly controlled geometric and radiometric calibration, high signal-to-noise ratios, and spectral alignment with the Sentinel and Landsat archives.
NASA's CAPSTONE concludes extended mission, validating autonomous optical navigation in cislunar orbit
On July 6, NASA reported that its CAPSTONE spacecraft concluded its extended mission in June 2026, closing out nearly four years of technology maturation in cislunar orbit. The microwave-sized satellite, owned and operated by Advanced Space and built by Terran Orbital, launched in June 2022 as the first U.S. commercial mission at the Moon. After a 15-month mission extension, it became a software-defined testbed for autonomous navigation, deep-space networking, and communications.
The autonomous Navigation, Guidance, and Control software (autoNGC) lets the spacecraft determine its position and trajectory without waiting for ground instructions. Portions had flown in Earth orbit before, but CAPSTONE marked the first test at the Moon. The test grew more demanding when Deep Space Network antennas were tied up supporting the Artemis II crewed test flight, cutting CAPSTONE's communications window to a few passes per week. Without regular data from Earth, autoNGC fixed the spacecraft's location using an onboard star tracker camera to image the Moon, Earth, and other celestial bodies. This optical navigation approach at times outperformed ground-based methods for real-time onboard navigation.
CAPSTONE also demonstrated delay/disruption tolerant networking (DTN), which stores data aboard the spacecraft when no link is available and forwards it automatically once contact resumes. It became the first spacecraft to fly the latest DTN protocols beyond Earth orbit and the first to run them in NASA's open-source core Flight System. In one demonstration, a transfer to Earth was interrupted mid-transmission, then resumed at the next opportunity with every piece of data arriving intact.
NASA's Research and Technology Mission Directorate managed the mission through the Small Spacecraft and Distributed Systems program at Ames Research Center, with the autoNGC and DTN demonstrations run by the SCaN Division at NASA Headquarters. NASA's activities on CAPSTONE ended in June 2026, though Advanced Space will continue operating the spacecraft as a technology development testbed.
NASA JPL's ERNEST rover completes 26 km autonomous desert traverse demonstrating next-gen Mars autonomy
The Exploration Rover for Navigating Extreme Sloped Terrain (ERNEST), a prototype developed at NASA's JPL, completed a 26 km (16 miles) autonomous traverse across a Southern California desert. The drive took more than 37 hours of driving time spread over seven days this past March, and ERNEST covered it almost entirely on its own, with what JPL described as minimal intervention from monitoring engineers.
The four-wheeled rover measures 1.2 meters long and drove at speeds up to 1 km/h. That pace far exceeds current planetary rovers: Perseverance took five years on Mars to cover roughly a marathon distance (26.2 miles, or 42.2 km). ERNEST's mobility comes from novel wheels and an active suspension system rather than the passive "rocker-bogie" design used on Perseverance and Opportunity. Instead of open pivot points distributing weight across six wheels, ERNEST uses two gimbaling joints on its front chassis to alter its gait, producing motions JPL likened to "squirming, wheel-walking, and obstacle-climbing," and each wheel can steer independently. Hari Nayar, lead principal technologist for the ERNEST team, noted that "while the rocker-bogie system has been very successful over the past 30 years, there's been a lot of research in that time on mobility and understanding terrain interaction."
The autonomy software was trained through reinforcement learning in a virtual environment, where the rover accumulated thousands of hours of experiential data within a few days by running many simulations in parallel. The team then validated it on an obstacle course at JPL's Mars Yard before moving to the desert, where tests included night and low-light driving to simulate lunar conditions.
Development of ERNEST began in 2022 with JPL internal research and development funding and has since moved under the NASA Science Mission Directorate's Exploration Science Strategy and Integration Office and the agency's Mars Exploration Program. NASA engineers intend ERNEST to serve as a model for larger, more capable rovers designed to travel farther and faster on the Moon and Mars.
MDA Space makes firm offer to acquire majority stake in AI-driven EO analytics firm CLS
On July 8, MDA Space Ltd. announced a firm and irrevocable offer to acquire an approximately 70% interest in Collecte Localisation Satellites (CLS), a Toulouse-based provider of AI-driven Earth observation value-added services and satellite IoT solutions. MDA Space would pay approximately €567 million (C$920 million) in cash to the shareholders of CLS's parent company, with France's space agency, the Centre national d'études spatiales (CNES), retaining the remaining roughly 30% stake.
Founded in 1986 as a CNES subsidiary, CLS employs about 1,200 people across 40 sites in 19 countries and serves more than 14,000 customers in approximately 150 countries. The company is expected to generate about €286 million (C$465 million) in revenue in 2026 (CAGR of 22% since 2023). The deal vertically integrates MDA's upstream satellite assets and near-real-time data services with CLS's downstream analytics, connectivity solutions, and proprietary algorithms, and adds a sales team of more than 100 to expand distribution of MDA's CHORUS SAR offering.
MDA Space expects the acquisition to double its recurring revenue stream and to be accretive to Adjusted EBITDA and Adjusted EPS within the first year of ownership, with CLS margins in line with its 2026 outlook of 18% to 20%. Should CLS be unable to refinance its existing debt at closing, MDA would fund approximately €198 million to retire it. The signing of a definitive agreement remains subject to French employee-representative consultation procedures and regulatory approvals, with completion expected by the end of 2026 or early 2027.
Government innovative EO demand: NGA seeks AI change detection, UK Space Agency opens £1.4M innovation call
On July 9, the National Geospatial-Intelligence Agency issued a Commercial Solutions Opening seeking commercial AI/ML technologies to automate global Foundation GEOINT change detection from satellite imagery. Tied to the agency's expanding Foundation Digital Twin initiative, the solicitation asks for automated global monitoring with quarterly worldwide change notifications, on-demand analysis of user-defined areas, detection of changes as small as 100 square meters, and continuous improvement of the models through AI and machine learning — the requirement is to flag where change has occurred, not to classify individual features.
A few days earlier, the UK Space Agency opened the Earth Observation Missions and Technology (EOMT) Innovation Call 1, offering up to £1.4 million in grants — up to £200,000 per project over a maximum of 12 months — for UK-based research and development on innovative EO technologies. The agency particularly encourages early Technology Readiness Level projects supporting climate monitoring, the environment, and dual-use and commercial applications.
Space4AI
CNN profiles Starcloud's $1.1B orbital AI data center plans ahead of Blackwell satellite
On July 7, CNN Business published a feature on Starcloud, the Y Combinator-bred company building AI data centers in orbit previously covered at Bits & Orbits. The piece profiles the firm at a $1.1 billion valuation following a $170 million Series A.
Starcloud launched the first Nvidia H100 GPU to orbit in November 2025. The company's pitch is running AI training and inference in low Earth orbit as terrestrial data centers hit power constraints. The feature previews Starcloud-2, a satellite carrying Nvidia's Blackwell GPU, planned for late 2026.
Environmental and scientific groups urge FCC to pause orbital data center applications
On July 9, a coalition of environmental and scientific groups asked the Federal Communications Commission to pause its processing of applications for orbital data center constellations.
The groups argue that the accelerating push to move computing infrastructure into orbit has raised the stakes in a wider campaign for stronger oversight of megaconstellations. The request would halt the FCC's review of the filings while those broader oversight questions remain unresolved.
Till next time,
Meta-beat Column of this week
Read also about the AI Pipeline that sits at the core, producing this Newsletter, including its ups and downs of this week: