Why do chips consume electricity? The physical process of computation involves transistors and interconnects repeatedly charging and discharging capacitances. CMOS dynamic power is often approximated as:<\/p>\n
P_dynamic \u2248 \u03b1CV\u00b2f<\/strong><\/p>\n The more circuitry that switches and the higher the operating frequency, the greater the power consumption generally becomes. Reducing voltage can significantly lower power consumption, but it also affects speed and operating margins.<\/p>\n Ten watts is not the power consumption of the entire system. An onboard inference operation must read sensor data and model weights, move them through memory, storage, high-speed buses, switches and interfaces, and then write or transmit the results. Error-correcting code, redundant execution and board-level DC\/DC conversion also consume power. The calculation must therefore include the processor, memory, storage, networking and conversion losses. The system should be evaluated by how many joules it consumes to complete one useful task, rather than solely by the chip\u2019s peak TOPS\/W.<\/p>\n That leads to two separate budgets. Power, measured in watts, determines whether multiple systems can operate simultaneously at a particular moment. Energy, measured in watt-hours and calculated as average power multiplied by operating time, determines whether they can continue operating throughout an orbital period. Spacecraft housekeeping computers generally remain online for long periods at relatively stable loads, while some payloads operate only during designated mission windows. If space-computing systems operate continuously at a high duty cycle, both average power and total energy demand rise.<\/p>\n ECSS requires average, peak and inrush power, eclipse periods\u2014when a low-Earth-orbit satellite enters Earth\u2019s shadow\u2014and fault conditions to be analyzed for each mission phase.[2] Consider an illustrative example: a low-Earth-orbit mission has a 90-minute orbit, including 60 minutes in sunlight and 30 minutes in eclipse. The spacecraft platform, computing, storage and communications systems continuously require 6 kilowatts. They therefore consume 9 kilowatt-hours per orbit, of which the battery must supply 3 kilowatt-hours during eclipse.<\/p>\n During the one hour of sunlight, the solar array must both supply the 6-kilowatt operating load and replenish the 3 kilowatt-hours consumed during eclipse. Even without losses, it must therefore generate 9 kilowatts. If regulation, charging, discharging and distribution losses are represented by an overall orbital efficiency of 85%, the solar array must provide at least approximately 10.6 kilowatts. If the battery has an allowable depth of discharge of 80% and a discharge efficiency of 95%, its nominal capacity must be at least approximately:<\/p>\n 3 \u00f7 (0.8 \u00d7 0.95) = 3.95 kilowatt-hours<\/strong><\/p>\n The 95% discharge efficiency is already included in the preceding 85% system efficiency and should not be counted twice.<\/p>\n NASA gives a solar irradiance of approximately 1,360.8 watts per square meter at 1 AU, while current multijunction space solar cells have nominal efficiencies of around 30% to 34%.[4][12] Assuming 30%, one square meter of active solar-cell area can generate no more than approximately 408 watts under normal incidence at beginning of life. The 10.6-kilowatt requirement in the example would correspond to approximately 26 square meters of active solar cells.<\/p>\n In practice, temperature, incidence angle, assembly gaps, shadowing, lifetime degradation and engineering margins must also be considered, making the complete array larger.<\/p>\n This is why space computing requires substantial energy. It is not that \u201cAI\u201d is inherently power-hungry, but that computing, memory and communications together create a relatively high and persistent load. A satellite must also generate enough electricity during its limited time in sunlight both to meet current demand and replenish the energy required during eclipse.<\/p>\n NASA\u2019s High Performance Spaceflight Computing system therefore incorporates fine-grained power controls that can shut down idle functions or place them in low-power modes.[3] In-orbit computing may also reduce total energy consumption by eliminating the downlink of unnecessary data. Ultimately, systems should be compared according to the end-to-end energy required to complete the same task.<\/p>\n Figure 1. Space-computing systems must balance both their power and energy budgets. Source: Author\u2019s illustration based on publicly available ECSS and NASA information.<\/em><\/p>\n Why do engineers begin discussing higher bus voltages as computing power increases? The answer lies in two fundamental equations:<\/p>\n P = VI<\/strong><\/p>\n P_loss = I\u00b2R<\/strong><\/p>\n Here, P is transmitted power, V is voltage, I is current, and R is line resistance. At the same power level, raising voltage reduces current. As current falls, I\u00b2R losses in harnesses, connectors and switching devices decline even faster.<\/p>\n Consider an idealized example intended solely to illustrate the relationship. Transmitting 5 kilowatts requires approximately 179 amperes on a 28-volt bus, but only 50 amperes on a 100-volt bus. If the two lines are provisionally assumed to have the same resistance, the latter\u2019s line loss is only about 7.8% of the former\u2019s.<\/p>\n Actual designs would not use identical wire gauges, topologies and converters. Nevertheless, these figures explain why high-power systems may exchange higher voltage for lower current, potentially reducing harness mass and voltage drop.<\/p>\n Figure 2. At the same power level, increasing bus voltage can reduce current and I\u00b2R line losses. The chart presents an idealized proportional example and does not represent the design of any specific spacecraft. Source: Author\u2019s illustration.<\/em><\/p>\n High-voltage risks fall into two categories. The first occurs inside equipment and during launch depressurization. When residual gas is present at certain combinations of pressure and gap distance, Paschen breakdown, corona or partial discharge may occur. Internal voids in materials and inadequately vented packages can also become weak points. NASA\u2019s current high-voltage design handbook, released in 2026, specifically addresses these risks.[5]<\/p>\n Verification focuses on insulation design, material and process controls, pressure sweeps, partial-discharge inception and dielectric-withstand testing.<\/p>\n The second category arises in the low-Earth-orbit plasma environment and involves spacecraft charging and discharging, solar-array parasitic currents, electrostatic discharge and arcing. NASA\u2019s low-Earth-orbit spacecraft charging handbook applies to LEO power systems operating above 55 volts and reviews the history of growing concern over solar-array arcing following the adoption of 100-volt systems.[6]<\/p>\n These risks must be analyzed in relation to orbital plasma conditions, array grounding, materials and exposed conductors. They must also be validated through current-collection and arcing tests in a plasma environment. Internal vacuum dielectric-withstand testing cannot serve as a substitute.<\/p>\n The opportunity created by high-voltage buses therefore extends well beyond selling a step-up converter. It simultaneously creates demand for high-voltage power devices, isolated gate drivers, low-parasitic packaging, high-voltage connectors and harnesses, insulating and potting materials, arc or abnormal-current detection, sectional isolation, and testing throughout the transition from atmospheric pressure through low pressure and into vacuum.<\/p>\n The optimal voltage is not necessarily the highest possible voltage. It is the point at which total system mass, total losses and risk-related costs are minimized.<\/p>\n Electricity generated by solar arrays cannot be delivered directly and unconditionally to computing boards. Between them sits the power conditioning and distribution unit, or PCDU\u2014or, more broadly, the power management and distribution system, or PMAD.<\/p>\n It must regulate generated power, stabilize the bus, manage battery charging and discharging, convert between voltage levels, switch loads, provide overcurrent protection, isolate faults and transmit status telemetry. NASA\u2019s 2026 review of small-spacecraft power systems also identifies conversion, power transmission and fault management as core PMAD functions. It notes that high-power payloads such as optical inter-satellite links and synthetic aperture radar have already expanded the power range these systems must handle.[4]<\/p>\n For computing satellites, the new challenge is that loads behave more like those in a data center than those of conventional instruments. Multiple computing nodes may raise or lower their power consumption on millisecond-to-second timescales. Input capacitors can create inrush currents during startup, while a short circuit on one board must not bring down the entire bus.<\/p>\n The PCDU must maintain bus stability while also knowing which loads must remain powered, which can be throttled and which must be disconnected immediately.<\/p>\n This creates three categories of opportunity. At the hardware level, they include high-efficiency DC\/DC converters, solid-state power controllers, latching current limiters and sectional isolation modules. At the architectural level, they include modular PCDUs that support parallel expansion and reuse across spacecraft models. At the software level, they include energy-management systems that jointly consider solar illumination, battery status, thermal margins and mission priorities.<\/p>\n However, \u201csoftware-defined power\u201d does not mean the protection chain can depend entirely on software. ECSS specifies that when battery energy falls to a defined threshold at which only critical loads and safe recovery can be supported, nonessential loads must be disconnected autonomously. The standards also require validation of inrush currents, undervoltage conditions and agreed representative faults in interactions between payloads and the main bus.[2][8]<\/p>\n An ESA PDU example similarly provides each load with independent latching current-limiter protection and acquires protection telemetry.[9] Hardware current limiting, independent shutdown and fault containment remain the foundation for safety-critical actions.<\/p>\n Figure 3. A PCDU connects power generation, energy storage, the main bus and different categories of loads, providing regulation, protection and telemetry. Source: Author\u2019s illustration based on publicly available NASA, ECSS and ESA information.<\/em><\/p>\n Modular PCDUs, high-efficiency power converters, solid-state power distribution and protection, high-voltage connections and insulation, energy-management software, and verification services covering inrush current, short circuits, undervoltage, electromagnetic compatibility, radiation and thermal-vacuum conditions.<\/p>\n The most obvious way to support continuous computing is to make the solar arrays larger. But a solar array is first a structural and mechanical system and only then a power-generating panel.<\/p>\n NASA\u2019s 2026 small-spacecraft technology review notes that deployed solar arrays are often a satellite\u2019s largest structures. Their dimensions and fundamental frequencies can affect attitude pointing, propulsion and orbit maintenance. The peak beginning-of-life power figures commonly reported by suppliers may also differ from both the power actually available to payloads and average power over an orbit.[4]<\/p>\n \u201cHigh power\u201d must therefore answer at least four questions:<\/p>\n Discussing only solar-cell efficiency overlooks the substrate, hinges, deployment actuators, cabling, locking mechanisms, thermal deformation and control-system coupling.<\/p>\n Roll-out arrays, flexible substrates and high-packaging-ratio designs deserve attention because they seek to reduce the need for heavy rigid panels and hinges. NASA information updated through June 2026 shows that the Roll-Out Solar Array, or ROSA, technology has been used for the International Space Station\u2019s upgraded solar arrays and the DART mission, and has also been incorporated into commercial geostationary satellite products.[10]<\/p>\n This demonstrates that the technology has achieved flight heritage and commercial adoption. It does not mean that it can be copied directly at any size or onto any platform. As an array grows, synchronized deployment, flexible vibration, cable routing and gravity-offload testing on the ground all become more difficult.<\/p>\n \u201cLow cost\u201d also requires a different denominator. Inexpensive solar cells may require more area, supporting mass and testing labor, leaving the overall satellite no cheaper. A more meaningful engineering metric is the cost of each watt reliably delivered to the bus over the mission lifetime.<\/p>\n That calculation includes the cells, complete array structure and mechanisms, conversion losses, launch mass, assembly and testing, production yield, lifetime degradation and failure risk.<\/p>\n The opportunities for cost reduction consequently become more specific: standardized array sections and mechanical-electrical interfaces, automated cell placement and inspection, repeatable deployment mechanisms, fewer parts and less one-off manual adjustment, volume production based on a common spacecraft platform, and the inclusion of end-of-life performance and structural-dynamics data in product specifications.<\/p>\n NASA\u2019s review notes that mass-production and automation capabilities for space solar arrays remain limited, while standardized constellation designs are becoming more common.[4]<\/p>\n Figure 4. High-power, low-cost solar arrays must be evaluated across electrical performance, structures, mechanisms, environmental resilience and manufacturing. Source: Author\u2019s illustration based on publicly available NASA information.<\/em><\/p>\n Solar arrays generate electricity only in sunlight. Energy storage must support low-Earth-orbit eclipse periods, attitude maneuvers, peak payload demand and emergency modes.<\/p>\n Battery sizing is constrained not only by watt-hours, but also by instantaneous discharge rate, temperature, allowable depth of discharge and cycle life. Repeated deep charging and discharging may reduce mass but accelerate aging. Oversizing the battery to reduce cycling, meanwhile, increases both launch mass and thermal-management requirements.<\/p>\n It is also important to avoid assuming that high power-system efficiency eliminates the thermal problem. Even a PCDU operating at 95% efficiency produces approximately 1 kilowatt of losses while processing 20 kilowatts. Most of the electricity consumed by computing equipment ultimately becomes heat as well.<\/p>\n High voltage reduces harness losses, but it does not eliminate heat from processors, storage devices, converters or batteries. Energy and thermal-control systems must therefore coordinate mission scheduling. Even when electricity is available, computing systems must still be throttled if the heat cannot be rejected.<\/p>\n Larger solar arrays also affect attitude control and communications. Flexible arrays continue to vibrate after maneuvers, while optical payloads and laser communications terminals require stable lines of sight. Controlling jitter may require slower maneuvers or waiting for deployed structures to settle before communications can begin.<\/p>\n Energy represents the first major opportunity precisely because it is the first subsystem to bring structures, thermal control, guidance, navigation and control, and mission operations onto the same trade-off table.<\/p>\n First, determine whether its metrics extend from components to the complete system. Power devices should provide efficiency data across the full input-voltage range, load range and temperature range. Solar-array suppliers should provide complete-array end-of-life power, specific power, stowed volume and fundamental frequency. PCDUs should provide data on inrush current, short-circuit isolation, bus recovery and telemetry accuracy.<\/p>\n Peak efficiency, beginning-of-life power or a photograph of a prototype alone cannot demonstrate delivery capability.<\/p>\n Second, determine whether verification covers the new failure modes. High-voltage systems require verification of insulation, partial discharge, material outgassing, low-pressure transitions, vacuum behavior, arcing and electromagnetic compatibility. Dynamic loads require realistic waveform waveform testing, concurrent startup, fault injection and degraded-mode operation. Large arrays require deployment, thermal-deformation, vibration and control-coupling tests.<\/p>\n Third, determine whether products can be reused across spacecraft models. If space computing develops into constellations, the energy system cannot be redesigned from scratch for every satellite. Scalable power modules, common telemetry protocols, standardized array sections, configurable protection thresholds and digital production records are necessary to spread development costs across a larger production run.<\/p>\n Past PCDU projects have already treated architectural optimization, testability and design-for-manufacturing rules as tools for reducing constellation costs. Specific products, however, must still be validated through current orders and in-orbit data.[11]<\/p>\n Fourth, determine whether customers are buying \u201crated kilowatts\u201d or \u201cavailable computing time.\u201d The deliverable that ultimately creates value is a guaranteed number of hours of computing and communications under specified orbital, lifetime, attitude and thermal conditions.<\/p>\n Companies capable of combining power generation, energy storage, distribution, heat rejection and scheduling into a verifiable service capability will be closer to becoming infrastructure suppliers for space computing.<\/p>\n In the age of space computing, the energy subsystem is both the first major opportunity and the first barrier because every ambitious computing narrative ultimately comes down to a simple chain of questions: Where does the electricity come from? How is it delivered? How are faults isolated? How does the system survive eclipse? And how is the resulting heat rejected?<\/strong><\/p>\n The most important signals to track next are not merely how many TOPS a computing satellite can deliver, but three harder indicators: whether end-of-life power and actual load profiles are disclosed; whether system-level validation of high voltage, dynamic power distribution and large-array deployment has been completed; and whether the same energy products are being delivered repeatedly across production batches and spacecraft platforms.<\/p>\n Only then will the opportunity move from conceptual exposure to genuine industrial capability.<\/p>\n![\"Space-computing](\"\/wp-content\/uploads\/2026\/09\/Space-computing-systems-must-balance-both-their-power-and-energy-budgets.webp\")
<\/p>\nHigh-Voltage Buses: Lower Current, More Complex Electrical Boundaries<\/h2>\n
![\"At](\"\/wp-content\/uploads\/2026\/09\/At-the-same-power-level-increasing-bus-voltage-can-reduce-current-and-I\u00b2R-line-losses.webp\")
<\/p>\nPCDUs: The Real Power-Dispatch Centers of Space Computing<\/h2>\n
![\"A](\"\/wp-content\/uploads\/2026\/09\/A-PCDU-connects-power-generation-energy-storage-the-main-bus-and-different-categories-of-loads-providing-regulation-protection-and-telemetry.webp\")
<\/p>\nSegments Most Likely to Secure Early Orders<\/h3>\n
High-Power, Low-Cost Solar Arrays: The Challenge Is Not Simply Adding More Cells<\/h2>\n
\n
![\"High-power,](\"\/wp-content\/uploads\/2026\/09\/High-power-low-cost-solar-arrays-must-be-evaluated-across-electrical-performance-structures-mechanisms-environmental-resilience-and-manufacturing.webp\")
<\/p>\nDo Not Overlook Batteries and Thermal Control: Power Generated Is Not the Same as Power Available<\/h2>\n
What Kind of Energy Company Is Truly Positioned to Capture the Opportunity?<\/h2>\n