SpaceX is pushing vertical integration deeper than rockets, satellites and launch infrastructure. Its own 2026 SEC materials describe Terafab as a SpaceX-Tesla initiative designed to produce one semiconductor class for terrestrial AI and a separate class optimized specifically for the space environment. CesiumAstro's August 2026 acquisition of RF-chip designer Jariet Technologies and Amazon's earlier Prometheus ASIC demonstrate that SpaceX is not acting in isolation: a silicon-to-satellite race is already underway, and three independent companies are assembling different pieces of the same macro pattern.
What Is SpaceX Actually Building With Terafab?
Terafab is a planned semiconductor manufacturing initiative involving SpaceX and Tesla, with Intel participating as an announced semiconductor-technology partner. Its intended scope extends unusually far across the semiconductor stack. SpaceX describes a closed-loop system spanning lithography-mask design, logic fabrication, memory fabrication and advanced packaging — the kind of integration that would give the company meaningful control over the silicon from design to finished device.
The significance is not merely that SpaceX wants custom chips. Technology companies have designed custom silicon for years. The notable development is the level of integration SpaceX is describing and, more importantly, the explicit connection its own filing makes between that integration and orbital computing. SpaceX describes a chip class built specifically for the space environment and specifically supporting SpaceX's orbital compute infrastructure. That connection did not require outside interpretation: SpaceX stated it directly to investors. The central thesis of the silicon-to-satellite argument is therefore unusually well-grounded — SpaceX has already made the case for its own vertically integrated orbital computing strategy in public regulatory filings.
How Much Is SpaceX Spending on Terafab?
The confirmed 2026 numbers require careful separation because three different figures have circulated around the project, and they originate from two different source documents at two different points in time. The current announced initial-phase investment is approximately $16.8 billion. Reuters and the Houston Chronicle reported that figure on August 6, 2026 in connection with the planned Grimes County, Texas complex, confirming both a planned footprint of roughly 100 million square feet and a commitment of at least 3,000 jobs. Those figures describe a planned complex; they should not be interpreted as evidence that the semiconductor operation is already built or producing chips.
Two much larger figures also exist. An earlier Grimes County local project proposal reported in May described approximately $55 billion as a proposed initial investment and approximately $119 billion as a potential full multi-phase buildout ceiling. Those numbers are real, but they did not originate in SpaceX's SEC IPO prospectus — a prior version of POPR's internal research graph incorrectly attributed them there, and that error has been corrected. The $55 billion and $119 billion came from the earlier local project proposal. The $16.8 billion came from the August 6 confirmed initial-phase announcement. The correct chronology is not that a $55 billion project shrank to $16.8 billion; they are different figures from different source records and different stages of the project, and future phases could eventually move total investment toward the earlier disclosed ceiling.
What Does the One-Terawatt Terafab Target Mean?
SpaceX has also stated a long-term Terafab goal associated with one terawatt of annual compute production. That figure requires precise handling. It is not a statement that the factory will continuously draw one terawatt from the electrical grid, it is not a description of current manufacturing output, and it is not a conventional semiconductor-fab capacity measurement such as wafers per month. It is SpaceX's stated long-term compute-production target, communicating the intended scale of the semiconductor program. Interpreting it as present electrical demand or current semiconductor production would convert a strategic target into a factual claim the evidence does not support.
Where Does Intel Fit Into SpaceX's Terafab Plan?
Intel has joined the project as a semiconductor-technology partner. SpaceX's SEC materials say Intel is expected to contribute expertise involving the design, fabrication and packaging of high-performance chips. The exact scope of Intel's role is not yet fully defined in the verified record, and the sealed evidence does not establish joint ownership of Terafab or co-development of every chip. The correctly evidenced formulation is that Terafab is a SpaceX-Tesla initiative with Intel participating as an announced semiconductor-technology partner. That distinction matters because semiconductor partnerships can range from consulting and process expertise to fabrication, packaging, intellectual-property licensing and deeper joint engineering — and which of those structures will ultimately define Intel's role has not yet been established.
Is SpaceX Trying to Stop Buying Chips From Outside Suppliers?
No. SpaceX's SEC materials explicitly state that the company expects to continue sourcing significant compute hardware from outside suppliers even after Terafab becomes operational. That makes Terafab a supplementation strategy rather than a complete supplier-replacement program, and it also makes engineering sense. Space systems contain many different semiconductor requirements. Some workloads benefit from highly specialized custom ASICs; others may require reconfigurable FPGAs; still others can continue using commercially sourced processors, memory and specialized components. Vertical integration does not require a company to manufacture every transistor in every subsystem — it requires the company to pull strategically important parts of the semiconductor stack closer to the mission architecture where doing so provides enough advantage to justify the investment.
Why Would SpaceX Want Chips Designed Specifically for Space?
Space is not simply another data-center environment. Electronics operating in orbit face a combination of constraints that terrestrial computing systems do not encounter. Radiation can degrade semiconductor devices through total ionizing dose accumulation. High-energy particles can trigger single-event effects. Power is constrained, heat rejection is difficult because spacecraft cannot rely on ordinary atmospheric convection, mass matters, physical volume matters, and a device may need to operate for years without physical servicing. These conditions affect semiconductor design decisions in ways that distinguish space-grade electronics from their terrestrial counterparts. A chip optimized solely for maximum performance per dollar may not be the ideal device for an orbital computing platform. The mission environment changes the silicon requirements, which is precisely why SpaceX's decision to distinguish terrestrial and space-optimized chip classes carries technical meaning: the space chip is not simply the terrestrial chip placed inside a satellite.
Is SpaceX Turning the Satellite Into a Computer?
That is increasingly the useful way to understand the direction of the architecture. A modern communications spacecraft already contains computing, networking, memory, radio-frequency electronics, antennas, power systems and software. The new development is the degree to which those elements can be co-designed when a company controls more of the underlying silicon. If a company controls more of the silicon, the radio architecture, the phased arrays, the networking software and the spacecraft itself, optimization can move across boundaries that previously belonged to separate vendors. The spacecraft becomes less like a vehicle carrying electronics and more like a distributed computing platform whose physical location happens to be orbit. SpaceX's own reference to orbital compute infrastructure makes that interpretation especially significant: the company is connecting its space-specific chip class directly to an orbital compute architecture, not describing space-qualified chips simply as components for conventional satellite missions.
Why the CesiumAstro Acquisition Matters
CesiumAstro announced its acquisition of Jariet Technologies on August 12, 2026 — two days before the final verification date for this report — and that transaction provides a second current example of semiconductor capability moving closer to a space-system company. But CesiumAstro and SpaceX are not doing the same thing, and that distinction is fundamental. SpaceX's Terafab program targets logic, memory and general-purpose compute semiconductor production at substantial scale, supporting both terrestrial and space-computing applications. CesiumAstro acquired a company specializing in RF and mixed-signal semiconductor design. Jariet is fabless and uses outside U.S.-based wafer fabrication facilities, so CesiumAstro acquired semiconductor design capability rather than semiconductor fabrication capability. That makes the acquisition a vertical-integration event without making it a fab acquisition — a distinction that matters when assessing what CesiumAstro actually gained.
What Does Jariet Technologies Actually Build?
Jariet was founded in 2015 and specializes in RF-sampling data converters and RF transceiver integrated circuits. Its Electra family reaches direct RF sampling up to 36 GHz with 64 GSPS conversion, according to Jariet's own materials and corroborating specialist reporting. That capability is particularly relevant to modern satellite communications because direct RF sampling can reduce the number of intermediate analog conversion stages between the antenna and the digital processing system. In a traditional radio architecture, incoming radio-frequency signals pass through multiple analog conversion stages before reaching a digital processor. Direct RF sampling moves the digitization boundary closer to the incoming radio signal, enabling more of the radio's behavior to be controlled through digital processing. For a company developing electronically steered phased arrays and software-defined communications payloads, owning more of the semiconductor technology underneath that signal chain can create tighter system-level optimization.
Why RF Silicon Is Different From SpaceX's Compute Chips
This distinction is one of the most important in the entire story. SpaceX's Terafab targets advanced logic, memory and general compute. CesiumAstro's Jariet acquisition targets RF sampling, mixed-signal conversion, transceiver technology and communications processing. One is aimed at the computational core; the other sits at the radio-frequency interface between the spacecraft and the electromagnetic environment. Calling both companies examples of custom chips in space is directionally correct but technically incomplete. The more interesting pattern is that different semiconductor layers are being pulled toward the companies designing the larger space system simultaneously — compute, memory, RF conversion, beamforming, networking, AI processing and packaging. The silicon-to-satellite race is therefore not one chip race; it is a race to control the semiconductor layers that most directly determine mission performance.
Why Does CesiumAstro Want Jariet In-House?
CesiumAstro described the acquisition as part of an effort to advance vertically integrated communications and ISR systems "from silicon to missions." A phased-array communications system is not simply an antenna. It combines semiconductor components, signal conversion, beamforming, digital processing, software and mission requirements. When these layers are developed by separate companies, each supplier optimizes around an interface. When more layers move under one architecture, optimization can span those interfaces, allowing chip requirements to derive from the radio and mission requirements instead of requiring the radio architecture to adapt to fixed commercial components. This is the same broad macro pattern visible at SpaceX, even though the specific semiconductor layer is different.
How Large Is CesiumAstro's Capital Expansion?
CesiumAstro's 2026 capital announcements are significant but easy to double-count, and the sealed POPR research specifically addresses the reconciliation. In January 2026, CesiumAstro announced a $200 million EXIM/J.P. Morgan financing package for its planned 270,000-square-foot Texas facility, while describing the broader Texas expansion as representing more than $500 million in capital investment over five years. In February, CesiumAstro announced $470 million in total growth capital — but that total consisted of $270 million in equity plus the same $200 million EXIM/J.P. Morgan financing already announced in January, not a second separate raise. In March, the company again described the headquarters and manufacturing expansion as a roughly $500 million five-year capital-investment program.
The correct non-duplicated statement is that CesiumAstro secured $470 million in growth capital in February, composed of $270 million in equity and $200 million in EXIM/J.P. Morgan financing, while separately describing a manufacturing expansion representing more than $500 million of capital investment over five years. Adding those figures together would count the same financing more than once. The acquisition of Jariet is occurring inside a much larger industrial expansion in which CesiumAstro is building a larger manufacturing and systems operation around the semiconductor design capability it now controls.
Did This Silicon-to-Satellite Trend Begin in 2026?
No. Amazon provides the clearest independent precedent. Amazon developed a custom ASIC called Prometheus for Project Kuiper, now renamed Amazon Leo, and public reporting on Prometheus dates to 2023 — roughly three years before the current SpaceX and CesiumAstro news cycle. The chip combines several functions that would traditionally be spread across separate components: a 5G modem, cellular base-station functionality and microwave backhaul, deployed across customer terminals, satellites and ground gateway antennas, with an architecture capable of processing up to 1 terabit per second per satellite. That historical timing matters: CesiumAstro's Jariet acquisition is not the origin of custom satellite silicon, and SpaceX's Terafab is not the first time a space-network company has designed silicon around its communications architecture. Amazon had already demonstrated the model. What changed in 2026 is the scale and breadth of the pattern.
Why Amazon's Prometheus Chip Is Important
Prometheus demonstrates that custom satellite silicon can be designed across the full communications system rather than only inside the spacecraft. The same chip architecture participates in customer terminals, satellites and gateways. That represents a strong form of vertical integration because the company can optimize both ends of the communications network together: the satellite is not designed separately from the ground terminal, the silicon is not designed separately from the network, and the network is not designed separately from the spacecraft. That is the same direction of travel visible in SpaceX's orbital-compute strategy and CesiumAstro's silicon-to-mission strategy, even as the companies pursue different markets and different semiconductor layers.
Why Project Kuiper Is Now Called Amazon Leo
Amazon officially renamed Project Kuiper to Amazon Leo on November 13, 2025. The Prometheus chip was publicly described before that rename, which is why older technical reporting correctly refers to Project Kuiper. The accurate 2026 formulation is therefore "Amazon's Prometheus chip, developed for Project Kuiper, now Amazon Leo" — preserving the historical source language while using the current product identity.
Is the Space Race Becoming a Semiconductor Race?
In part, yes — but that conclusion needs to remain proportional to the evidence. Launch capability still matters, satellite manufacturing still matters, and spectrum, ground infrastructure, power systems, antennas and software all remain central to the space competition. What has changed is that semiconductor architecture increasingly determines what those systems can do. Compute performance affects onboard processing; RF conversion affects radio architecture; beamforming affects phased-array performance; power efficiency affects spacecraft size and thermal design; radiation tolerance affects reliability; memory affects local data handling; and AI acceleration affects what can be processed onboard instead of transmitted elsewhere. A company that controls more of those semiconductor decisions can potentially optimize the complete mission around them, which makes silicon a deeper strategic layer of the space competition than it was when most spacecraft electronics were assembled primarily from standardized external components.
Why Phased Arrays Make Custom Silicon More Important
Electronically steered phased arrays are especially dependent on tight coordination between hardware and signal processing. A phased array contains many antenna elements whose signals have to be controlled together to steer and shape a beam electronically, requiring precise timing, data conversion, RF processing and digital control. As arrays scale and frequencies rise, the relationship between the antenna, the RF silicon, the converter architecture and the digital processor becomes increasingly consequential. CesiumAstro's acquisition of Jariet is therefore not simply a semiconductor diversification move. It brings direct RF-sampling expertise closer to a company already building active phased-array communications systems, enabling the chip to increasingly be designed around the array and the array to increasingly be designed around the mission — vertical integration at the radio layer.
Why AI Changes the Satellite Architecture
AI introduces another pressure toward onboard computing. A satellite that can process more information locally does not have to transmit every raw data stream to the ground before useful decisions can occur, and that matters for communications, sensing and ISR systems alike. The verified graph does not establish that every satellite will become an autonomous AI system or that terrestrial cloud computing will be displaced by orbit. The stronger conclusion is structural: as onboard processing becomes more important, the performance, power efficiency, memory and radiation tolerance of space-grade compute become more consequential. That makes SpaceX's intention to build a space-specific chip class for orbital compute architecturally significant in a way that extends beyond the satellite manufacturing story — the company is extending vertical integration into the semiconductor architecture underneath orbital computation itself.
Why Space Chips Cannot Simply Follow Terrestrial AI
The terrestrial AI industry currently rewards enormous compute density — the biggest, fastest processor often wins the benchmark. Space imposes additional constraints that change which design is optimal. The chip has to survive radiation, the spacecraft has limited power, heat has to be rejected without ordinary atmospheric cooling, mass launched into orbit carries real economic cost, and hardware replacement can be impossible or prohibitively expensive once the spacecraft is deployed. Those conditions mean the largest or fastest terrestrial processor is not automatically the best space processor. Custom silicon becomes strategically valuable because a company can optimize for the actual mission envelope instead of accepting all the assumptions built into a component designed for a very different environment.
Does Custom Silicon Mean FPGAs Are Going Away?
No. The verified engineering outlook supports a heterogeneous architecture in which FPGAs remain valuable because they can be reconfigured after manufacturing and can support mission-specific processing without requiring a new custom chip for every function. ASICs offer advantages when workloads are stable enough and scale is large enough to justify custom design costs. The realistic near-term architecture is therefore not ASICs replacing FPGAs — it is custom ASICs, FPGAs and externally sourced semiconductor components coexisting, with each used where its characteristics fit the mission. SpaceX's own continued third-party sourcing plans reinforce the same conclusion. The silicon-to-satellite race is not a march toward total semiconductor self-sufficiency; it is a move toward selective control over the chips that matter most to mission performance.
What SpaceX's Strategy Could Mean Two or Three Years From Now
The immediate story is a factory announcement. The longer-term story is architectural leverage. If Terafab progresses toward its intended role, SpaceX could increasingly co-design orbital compute hardware around the Starlink and broader SpaceX system rather than treating silicon as an external constraint — affecting power efficiency, thermal architecture, onboard AI, networking, memory, radiation strategy, packaging and potentially the way future orbital computing platforms are structured. The important evidence boundary is that Terafab is still a planned and developing project. The facility is not yet producing its stated future output, the $16.8 billion figure describes an announced initial investment and the terawatt-scale compute target is a long-term goal. The architectural significance comes from what SpaceX says it intends to build, not from treating those capabilities as already operational.
Why the Commercial Stakes Are Larger Than Satellites
Terafab also sits at the intersection of two enormous technology programs simultaneously. One semiconductor class is intended for SpaceX's orbital compute infrastructure; the other is intended for Tesla's terrestrial edge and inference workloads, including Tesla vehicles and Optimus. That makes the project unusual even within vertical semiconductor integration: the same manufacturing architecture is intended to support autonomous vehicles, robotics and space computing. The workloads are different and the environmental requirements are different, but the strategic motivation is shared. Compute has become sufficiently important to the larger system that the companies want greater control over its foundational hardware. That is why Terafab is not simply a SpaceX satellite-supply story — it is a SpaceX-Tesla semiconductor-platform story whose scope extends from the factory floor to the edge of the atmosphere and beyond.
The Silicon-to-Satellite Race Is Really a Co-Design Race
The most useful way to interpret the three confirmed examples is not simply that more space companies are designing chips. Amazon developed Prometheus around the complete broadband architecture, optimizing both ends of the network from the same silicon. CesiumAstro is pulling RF-sampling semiconductor expertise closer to its phased-array and mission systems. SpaceX intends to manufacture space-optimized silicon for an orbital compute architecture while retaining control over many of the surrounding systems. The common pattern is co-design: mission requirements influence chip requirements, chip capabilities influence payload architecture, payload architecture influences software, software influences network behavior and network behavior influences the mission. When enough of those layers sit inside one organization or a tightly coordinated system, the old boundaries between spacecraft, computer, radio and network begin to dissolve. That may be the deeper significance of the 2026 space-semiconductor story. The satellite is increasingly becoming a vertically integrated computer — and the companies that get there first will define what the next generation of orbital infrastructure looks like.
Fact Summary
Is SpaceX building its own semiconductor factory? SpaceX's SEC registration and prospectus materials describe Terafab as a SpaceX-Tesla vertical-integration initiative spanning lithography-mask design, logic fabrication, memory fabrication and advanced packaging.
How much is SpaceX initially investing in Terafab? The confirmed August 6, 2026 initial-phase investment is approximately $16.8 billion, reported by Reuters and the Houston Chronicle in connection with the planned Grimes County, Texas complex.
What are the $55 billion and $119 billion Terafab figures? They came from an earlier Grimes County project proposal reported in May 2026. The $55 billion was a proposed initial investment and the $119 billion was a potential full multi-phase buildout ceiling. Neither figure originated in SpaceX's SEC IPO prospectus.
How large is the planned complex? Roughly 100 million square feet and at least 3,000 planned jobs, per Reuters and Houston Chronicle reporting tied to the August 6 announcement.
What chips does SpaceX intend to make? One semiconductor class for terrestrial edge and inference use serving Tesla vehicles and Optimus, and another class optimized specifically for the space environment and SpaceX's orbital compute infrastructure.
Is Intel a co-owner of Terafab? That has not been established. Intel is an announced semiconductor-technology partner expected to contribute design, fabrication and packaging expertise, but its precise role remains incompletely defined.
Will SpaceX stop buying outside chips? No. SpaceX's SEC materials state the company expects to continue sourcing significant computing hardware from outside suppliers even after Terafab becomes operational.
What did CesiumAstro acquire? CesiumAstro acquired Jariet Technologies on August 12, 2026, a fabless RF-semiconductor company specializing in direct RF-sampling data converters and RF transceiver ICs.
Did CesiumAstro acquire a semiconductor fab? No. Jariet uses outside U.S.-based wafer fabrication facilities. CesiumAstro acquired semiconductor design capability, not fabrication capability.
How fast is Jariet's verified RF-sampling technology? Jariet's Electra family is publicly described at up to 36 GHz with 64 GSPS conversion.
How much capital has CesiumAstro raised? The February 2026 growth-capital package totals $470 million, consisting of $270 million in equity and the same $200 million EXIM/J.P. Morgan financing announced in January.
What is the $500 million CesiumAstro figure? It describes the company's planned Texas capital-investment program over five years, not an additional financing round. Adding it to the $470 million would double-count the EXIM financing.
Did custom satellite silicon begin in 2026? No. Amazon's Prometheus ASIC was publicly documented in 2023 for Project Kuiper, now Amazon Leo, establishing the precedent roughly three years before the current news cycle.
What does Amazon's Prometheus chip do? It combines 5G modem, cellular base-station and microwave-backhaul functionality, deployed across customer terminals, satellites and gateways, processing up to 1 terabit per second per satellite.
Is space moving entirely to custom ASICs? No. The verified engineering outlook supports a heterogeneous architecture combining custom ASICs, FPGAs and externally sourced components.
What is the larger silicon-to-satellite trend? Space companies with sufficient scale are pulling mission-critical semiconductor design closer to spacecraft and network architecture so chip requirements can increasingly derive from mission requirements rather than from available commercial silicon.
Evidence Status
CONFIRMED: SpaceX is the primary commercial anchor of the silicon-to-satellite story because its own SEC materials explicitly connect Terafab to a space-optimized semiconductor class supporting orbital compute infrastructure.
CONFIRMED: The current announced initial Terafab investment is approximately $16.8 billion, with a planned footprint of roughly 100 million square feet and at least 3,000 jobs.
CONFIRMED: The earlier $55 billion and $119 billion figures derive from a separate Grimes County project proposal and must not be attributed to SpaceX's SEC IPO prospectus. This was a real sourcing error in POPR's prior internal research, corrected per hostile audit.
CONFIRMED: Intel is participating as an announced semiconductor-technology partner, but the verified evidence does not establish joint ownership or fully defined co-development responsibilities.
CONFIRMED: SpaceX expects substantial outside semiconductor sourcing to continue alongside Terafab.
CONFIRMED: CesiumAstro acquired Jariet Technologies on August 12, 2026, bringing fabless RF-semiconductor design capability in-house.
CONFIRMED: Jariet's independently supportable direct RF-sampling capability reaches 36 GHz with 64 GSPS conversion.
CONFIRMED: CesiumAstro's $470 million February capital package includes the previously announced $200 million EXIM/J.P. Morgan financing. The separate $500 million figure describes planned five-year capital investment and must not be added as another financing round.
CONFIRMED: Amazon's Prometheus ASIC predates the 2026 news cycle and establishes a historical precedent for vertically integrated satellite silicon.
CONFIRMED: Space electronics face radiation, thermal, power and mass constraints that make semiconductor optimization materially different from terrestrial computing.
SUPPORTED MACRO CONCLUSION: Space-system companies with sufficient scale are increasingly pulling mission-critical semiconductor capability closer to spacecraft and network architecture.
SUPPORTED INTERPRETATION: The satellite is increasingly becoming a vertically integrated computing platform rather than simply a spacecraft carrying electronics.
NOT ESTABLISHED: That SpaceX intends to replace every external semiconductor supplier; that Terafab is already operational; that its one-terawatt compute-production target represents grid draw or current output; that Intel is a co-owner of Terafab; that CesiumAstro acquired semiconductor fabrication capability; or that every future satellite architecture will move exclusively to custom ASICs.
Sources
- SpaceX. SEC registration and prospectus materials. Terafab vertical-integration strategy, dual chip-class architecture (terrestrial and space), continued third-party sourcing intent, Intel participation description.
- Reuters. August 6, 2026 reporting on the approximately $16.8 billion initial Terafab investment, planned Grimes County complex and jobs commitment.
- Houston Chronicle. August 2026 independent reporting corroborating the $16.8 billion initial phase, planned 100-million-square-foot footprint and 3,000-plus jobs.
- TechCrunch. May 2026 reporting tracing the approximately $55 billion initial proposal and approximately $119 billion potential multi-phase ceiling to the earlier Grimes County project proposal.
- CesiumAstro. "CesiumAstro Acquires Jariet Technologies to Advance Vertically Integrated Communications and ISR from Silicon to Missions." August 12, 2026.
- Jariet Technologies. Corporate and product documentation confirming fabless status, outside U.S.-based wafer fabrication, RF-sampling specialization and Electra-family performance up to 36 GHz and 64 GSPS.
- CesiumAstro. January 15, 2026 announcement of the $200 million EXIM/J.P. Morgan financing package for the 270,000-square-foot Texas facility and the broader $500-million-plus five-year expansion.
- CesiumAstro. February 2, 2026 announcement of $470 million in total growth capital consisting of $270 million in equity and the previously announced $200 million financing package.
- CesiumAstro. March 6, 2026 description of its approximately $500 million five-year headquarters and manufacturing capital-investment program.
- Amazon. Public technical documentation describing the Prometheus custom ASIC for Project Kuiper, combining modem, base-station and microwave-backhaul functionality across terminals, spacecraft and ground gateways, processing up to 1 Tbps per satellite.
- Amazon. November 13, 2025 announcement renaming Project Kuiper as Amazon Leo.
- NASA and ESA. Established spacecraft-electronics literature on total ionizing dose, single-event effects and the power, thermal, mass and radiation constraints affecting space-grade electronics.