The Foundry Crunch That's Coming
Rad-Hard Chip Capacity and the 2026–2028 Constellation Ramp
What This Means
Three major U.S. government satellite programs, the Space Development Agency’s (SDA) proliferated low-Earth orbit (pLEO) architecture, the Space Force GPS III Follow-On (GPS IIIF), and NRO next-generation systems, are ramping concurrently between 2025 and 2028 and drawing from the same small pool of Defense Microelectronics Activity (DMEA) Trusted Foundry-qualified fabricators. Fewer than five domestic foundries hold the qualifications needed to produce radiation-hardened (rad-hard) semiconductors for these programs. The capacity math does not close. The deeper problem is structural: no Defense Federal Acquisition Regulation Supplement (DFARS) clause requires prime contractors to disclose foundry booking conflicts to government program offices, so the risk will not surface until it appears in a delivery slip. Program managers who do not act before that moment are already late.
The Foundry Layer Nobody Is Watching
Every classified and unclassified satellite in a U.S. government constellation depends, at some point in its build, on a radiation-hardened semiconductor fabricated at one of a small number of domestically qualified foundries. These are not interchangeable with commercial fabs. A rad-hard processor, memory device, or power management chip must be fabricated on a process specifically qualified for space radiation environments, and that qualification lives at the foundry, not the chip designer. Moving production to a different fab means requalifying from scratch, at timelines that routinely run 18 to 36 months.
The DMEA Trusted Foundry Program maintains the official list of U.S.-based foundries authorized to fabricate such components for government programs. Among the publicly confirmed participants: Microchip Technology (which absorbed Microsemi and its Actel heritage product lines), BAE Systems Electronic Systems, Honeywell Aerospace, and SkyWater Technology. The full roster may include additional participants not publicly disclosed, but the scale of the qualified domestic base is, by any measure, narrow.
These foundries operate primarily on process nodes ranging from approximately 90 nanometers (nm) to 350 nm, the range reflecting the long qualification timelines and capital requirements specific to rad-hard processes. SkyWater Technology has publicly confirmed its 90 nm and 130 nm rad-hard process capabilities in its filings with the Securities and Exchange Commission (SEC). Process-node characterizations for BAE Systems and Honeywell rest on trade-press reporting at the Class 2 level and should be read as context rather than confirmed operational specifications. What matters for this analysis is the structural reality: commercial leading-edge fabs are building at 3 nm to 5 nm. The domestic rad-hard foundry base is operating between one and four generations behind, not because of negligence, but because radiation-hardened process qualification demands a stability that commercial scaling does not.
Allied-nation qualified foundries are outside the scope of this analysis. The DMEA Trusted Foundry Program is a U.S.-specific qualification framework, and no allied-nation equivalent qualification is currently cross-accepted for U.S. government satellite programs under existing International Traffic in Arms Regulations (ITAR) and Export Administration Regulations (EAR) constraints. European suppliers such as STMicroelectronics produce radiation-tolerant components and serve commercial and some allied programs, but they do not hold DMEA Trusted Foundry qualification and cannot substitute for domestically qualified production on the programs discussed here. The capacity question is therefore bounded to the domestic DMEA-qualified base.
Three Programs, One Foundry Pool
The demand problem is not a future scenario. It is a present condition with a fixed horizon.
SDA pLEO Tranche 1 and 2. The Space Development Agency has awarded Tranche 1 contracts to multiple prime contractors; those satellites are in delivery now. Tranche 2 contracts have been awarded to Northrop Grumman, York Space, and others, with production underway. Tranche 3 solicitation activity is documented on SAM.gov. The SDA architecture calls for hundreds of satellites across successive tranches, each requiring rad-hard and radiation-tolerant semiconductors for processing, communications, and power management subsystems. The 2024 through 2027 delivery concentration for Tranche 1 and 2 is confirmed in SDA contract documentation and corroborated by Government Accountability Office (GAO) program assessments.
GPS IIIF. The GPS III Follow-On contract with Lockheed Martin Space is a matter of public record via the Department of Defense (DoD) and Space Force budget exhibits. The program includes a production run of satellites with deliveries planned through the late 2020s. “Planned” is the operative word, future-dated milestones can shift, and readers should treat schedule projections as subject to program execution risk. What is not in dispute is the sustained nature of the demand: GPS IIIF is not a one-time procurement but a multi-year draw on the same qualified component base.
NRO next-generation systems. Program specifics are classified and will not be characterized beyond what unclassified budget exhibits and congressional authorization documentation support. What those sources confirm is that NRO next-generation satellite procurement is active during the same 2025 through 2028 window. The NRO demand stream is additive to the capacity calculation, and it is the one demand stream that program managers at SDA or GPS program offices have no visibility into. NRO-specific chip demand cannot be quantified from unclassified sources; its magnitude relative to SDA and GPS IIIF is not assessable from public information. It is characterized as additive and concurrent, not sized.
The convergence is the story. Three program families, each with independent program offices, each managing delivery risk against their own prime contractor commitments, all drawing from the same four-foundry pool, with no mechanism in place to tell any of them that the other two are in the queue ahead of them.
The Capacity Math
The “capacity math does not close” claim in the Signal Summary deserves a structural demonstration, even at the inference level. Here is the public-record basis for it.
SDA’s Tranche 1 called for 28 transport-layer satellites and 10 battle management satellites across its initial awards, with Tranche 2 scaling to roughly 100 transport-layer satellites across multiple prime awards. SDA has stated publicly that its architecture scales to hundreds of satellites across successive tranches. Each satellite carries multiple rad-hard or radiation-tolerant integrated circuits across its processing, communications, and power management subsystems; industry-standard satellite bus architectures for this class of vehicle typically incorporate dozens of rad-hard components per platform. The chip-unit demand across SDA Tranche 1 and 2 alone, at even a conservative estimate of 20 to 30 rad-hard components per satellite, produces a demand figure in the range of several thousand chip-units against a foundry base that, by confirmed participant count, numbers fewer than five domestic fabs.
GPS IIIF adds a sustained multi-year draw from the same foundry pool. Its production quantities are documented in DoD budget exhibits; the program is not a large-volume procurement by commercial standards, but it draws on the same small set of qualified process nodes and competes for the same foundry capacity windows.
No public source discloses foundry throughput, utilization rates, or booking queue data for any DMEA Trusted Foundry participant. The capacity gap is therefore a structural inference, not a measured shortfall: a confirmed small foundry pool, confirmed concurrent demand across three independent program families, and no disclosed mechanism for cross-program allocation visibility. The inference does not require actuarial precision to be decision-relevant. If any one of the three programs has assumed foundry availability that one of the other two has already booked, the slip will not be visible until it is a delivery problem. That is the operational meaning of “the capacity math does not close.”
The next section maps the specific sub-tier fabless designers where the allocation conflict will first appear, names the prime-to-sub-tier contractual gap that leaves program offices flying blind, and documents the specific DFARS provisions that currently fail to require foundry disclosure. Paid subscribers also receive the full five-action decision framework structured by audience role, the CHIPS Act timeline analysis, and the 18-month window calculation for alternative-source qualification. This is the supply chain map your program office or investment committee needs before the slip becomes visible.
