Ask most people what a space program is and they describe a satellite. Ask a program manager where the schedule went wrong and the answer is often on the ground.
The ground segment is the antennas, modems, scheduling software, mission control, and the operators trained to use them. It is less photogenic than a launch. It also fails slowly and expensively, and the public record from the last few years shows how.
GPS: a ground system that ran years late, then was cancelled
The clearest example is OCX, the ground control system for modern GPS satellites. Defense News reported its cost estimate rising from $3.7 billion in 2012 to $4.3 billion in 2015 and $6.2 billion in 2018. In June 2023, GAO blamed the latest slip largely on “overlapping efforts” to test satellite capabilities and integrate software.
The satellites did not wait politely. The same article notes that OCX Block 0, delivered in 2017, could support GPS III launches but could not operate satellites in orbit.
The Pentagon’s test office raised a related point. As summarized by Breaking Defense, its FY23 report said there had not been enough testing of how satellites, ground control, and receivers work together. It also stated: “No operational testing was conducted in FY23 across the GPS Enterprise.”
Space Force accepted OCX from Raytheon in July 2025 and began integrated testing with the wider GPS enterprise, according to the Space Force’s own announcement. Acceptance did not mean operational, and it turned out not to be the finish line. On April 17, 2026, the Defense Acquisition Executive cancelled the program. Breaking Defense reported that officials judged OCX could not deliver needed capabilities on an operationally relevant timeline at acceptable risk, after extensive system issues surfaced in integrated testing. As of January 2026, the Space Force put the program cost at about $6.27 billion. It said it would keep improving the existing ground system, the Architecture Evolution Plan, rather than put more money into OCX.
Correction, October 21, 2026: an earlier version of this post described OCX as accepted in July 2025 with operations to follow and did not mention the April 2026 cancellation. It has been updated.
The lesson is not that one contractor struggled. A ground system is software tied to hardware, operators, and satellites. When those pieces are tested late and in overlap, the schedule has nowhere to absorb the surprises. In OCX’s case, the problems that late integrated testing exposed were serious enough to end the program after 16 years.
Capacity is a second ground risk
Even a ground system that works can run out of room. GAO-23-105505, published in April 2023, looked at the Satellite Control Network, the 19-antenna system that supports launch and daily control of many government satellites.
Utilization had averaged 75 percent over a decade, above the 70 percent level officials cite as the signal to add capacity. Annual launches it supports have tripled since 2012, and the network handles more than 450 contacts a day. Satellite users told GAO that limited availability could compromise their missions. Space Force asked users to cut non-critical contacts while it planned 12 new antennas and looked at commercial and other federal options.
GAO also found the network’s sustainment plan dated from 2017 and did not cover the current upgrade efforts. Space Systems Command approved an updated plan in August 2023. I have not checked where the new antennas stand today.
Buying contact time instead of building antennas
Commercial ground stations as a service change the question. One overview describes the model as turning antenna access into a schedulable operating expense: you buy contacts, often with data delivered into a cloud account, rather than building a site. Providers it names include AWS Ground Station, KSAT, Leaf Space, and ATLAS Space Operations.
It also lists the limits. A low Earth orbit satellite is over any one antenna for only a few minutes. Shared antennas can have scheduling conflicts. Not every station supports every frequency band. Lock-in, cybersecurity, and data sovereignty remain open concerns. The same piece notes reports that Microsoft retired Azure Orbital in 2025, a reminder that this market has churn.
What a program manager can do about it
These are opinions, drawn from the record above rather than from any one program.
- Give the ground segment its own owner and schedule. If it is a line item under the satellite, it will be managed as an afterthought until it is the critical path.
- Test end to end early. Waiting for flight hardware to exercise the ground chain repeats the OCX pattern. A simulated or emulated RF link lets operators rehearse commands, handovers, and failure modes before launch.
- Keep the radio layer software-defined where you can. If a band or modulation change means new hardware, you have turned a configuration problem into a procurement problem.
- Define the interface, not the station. If the contract is written around contact schedules and data delivery, you can move between your own antennas and commercial ones. If it is written around one vendor’s console, you cannot.
- Schedule operator training and readiness exercises as real work. The OCX record shows training on the old system while building the new one is a constraint, not a footnote.
The takeaway
Satellites get designed carefully because failure is visible and unrepairable. Ground systems are repairable, so they tend to get less rigor early and more pain later. Treat the ground segment as a first-class part of the architecture, test it with the rest of the system, and keep it swappable. That is cheaper than discovering at launch that the satellite is ready and the way to talk to it is not.
Stellabyte LLC is an AI, RF systems, and technology consulting firm based in California. Get in touch if you’re working on a problem that needs domain expertise, not just software.