Space-Based Solar Power

Space-based solar power is the proposal to collect sunlight in orbit, where there is no night, no weather and no atmosphere, convert it to a microwave or laser beam, and send it to a receiving antenna on the ground. The physical attraction is real and easily stated: a satellite in geostationary orbit sees the Sun more than 99% of the year, so it delivers something a terrestrial solar farm cannot deliver at any price — firm, dispatchable power from a solar resource, without storage. Every serious proposal since Peter Glaser's in 1968 is a variation on that one sentence.

The difficulty is equally easily stated, and it is arithmetic rather than physics. Nothing about the concept violates any law. What it requires is a structure hundreds of metres to a kilometre across, assembled robotically in geostationary orbit from something between one and two million modules, weighing between two thousand and eighty thousand tonnes depending on whose estimate you take, launched at a price per kilogram that has never been charged, beaming through an aperture large enough that the ground station occupies a permanently fenced site of thirty-four to a hundred square kilometres. Each of those requirements is individually conceivable. Nobody has demonstrated any of them above laboratory scale.

This brief is written around one measurement, because the measurement is what the subject's literature most consistently omits. The only spacecraft that has ever aimed a power beam at the Earth — Caltech's Space Solar Power Demonstrator, in 2023 — produced at the receiving dish a signal that its own builders classify, using their own published definitions, as a detection and not a power transfer, because the recovered power was less than the power the receiver needed to operate. That is not a criticism of the experiment, which was well designed and honestly reported. It is the reason a gigawatt-class solar power satellite is an extrapolation from a design rather than from a demonstration, and the reason this brief distinguishes demonstrated from designed from assumed in every paragraph that carries a number.

The scope boundary with the sibling slot. Wireless energy transmission owns power beaming that stays inside the atmosphere or close to it: inductive charging, electrified roadways, laser links across a range, the terrestrial and aerial microwave record. This brief owns the orbital segment — collection in space, the space-to-ground link over 35,786 km, the mass and launch economics of putting a power station in geostationary orbit, and the ground station that receives the beam. Where the two meet, at rectenna physics and at human-exposure limits, they cite the same sources and split by question. The rule of thumb is altitude: if the transmitter is in orbit, it is here.

One consequence of that cut is worth stating at the outset. The best measured far-field microwave power transfer anywhere on Earth — 1,180 W over just above 100 m at 20.8% DC-to-DC, by Xidian University in May 2026 — belongs to the sibling brief as a terrestrial result. It is nonetheless the most relevant single number in this one, because it is the state of the art in the conversion chain that a solar power satellite depends on, and it was achieved at one ten-thousandth of the range.

Established The 2023 Caltech demonstration delivered a detection to Earth, not power, and its authors say so explicitly. The Space Solar Power Demonstrator launched on 3 January 2023 into a 527 km, 97.5° orbit and operated for about ten months. Its beaming experiment, MAPLE, was a 32-element phased array of co-cured popup dipoles in two sixteen-element sub-arrays, roughly 160 × 90 mm, radiating at 9.984 GHz. Into rectennas mounted inside the enclosure, of order 30 cm away, it delivered a peak rectified 251 mW broadside and 231 mW at the sidewall, at a rectenna RF-to-DC efficiency of 40% with respect to the power incident on its aperture. Its equivalent isotropic radiated power was between 26.5 and 28.5 dBm — under a watt.

Established Then it pointed at the ground. Free-space path loss to the Caltech campus was 167 dB, the predicted signal −138.5 to −140.5 dBm, and the receiver a 67 cm dish of 33.5 dBi gain on the roof of the Moore Laboratory in Pasadena. After two unsuccessful attempts in May and June, a pass on 28 July 2023 produced a signal with the predicted Doppler shift. The paper's own definitions are the point: power transfer is “defined as a net positive recovery,” while detection is “a process where the recovered power is less than the external power needed for the receiver.” The authors classify the Earth experiment as detection. They also explain why the result was below even the modest budget: the ground station “was never within MAPLE's −3dB contour, due to the host's inability to maintain alignment.”

Established Caltech's own press release said “detectable power.” Much of the coverage did not. The June 2023 institutional announcement stated that MAPLE had “demonstrated its ability to wirelessly transmit power in space and to beam detectable power to Earth for the first time,” with no distance, no wattage and no receiver detail. Seven months later a widely read trade account reported that MAPLE “successfully demonstrated for the first time ever that solar power can be collected by photovoltaic cells and transmitted down to Earth via a microwave beam.” That sentence is not supported by the primary paper. This brief reports the conflict and sides with the primary source, and it notes that the institution was careful and the coverage was not.

Established The scale gap should be stated as a ratio, because a ratio is the only honest summary. Two hundred and fifty-one milliwatts at thirty centimetres against one gigawatt is a factor of about 4 × 10⁹. The space-to-Earth measurement, taking the v0.17 research pack's figure of roughly one femtowatt, against one gigawatt, is 10²⁴. The second ratio is the one that matters, because it measures the function the technology is named for. The arithmetic is this brief's; the inputs are sourced.

Established MAPLE also degraded, and the degradation was not anticipated. Averaged over eighteen benchmark measurements, transmitted power fell 14.7% over the eight-month test campaign, with peak achievable power down about 23%, attributed to failure of individual transmitting elements and to electrical–thermal interactions. The team's stated priority for improvement is thermal design. In a system whose reference designs assume decades of unattended operation on a structure of millions of elements, a 15% loss in eight months from element failure is a first-order finding rather than a footnote.

Frontier The structural demonstration is the one that should worry a reader most, and it is the one least discussed. DOLCE, the deployable-structure experiment, was 1.8 m × 1.8 m. Its intended three-to-four-day deployment stalled when a wire snagged and damaged a boom connection, and later part of the structure jammed under the deployment mechanism — a failure mode, in Caltech's own account, never observed in laboratory testing. Both were resolved on orbit. The programme's summary is that the test “demonstrated the robustness of the basic concept,” which is fair. The other fair reading is that 1.8 m surprised an expert team twice, and the reference designs need hundreds of metres to a kilometre.

Established The cost estimates in circulation span more than fiftyfold, and the widest spread is inside a single report. NASA's Office of Technology, Policy and Strategy published baselines of $610/MWh for its heliostat-swarm design and $1,590/MWh for its planar-array design, which it described as “12–31 and 32–80 times higher” than 2050 terrestrial projections. The same report's sensitivity analysis, substituting electric propulsion, lower launch cost and longer component life, reaches $200/MWh and $500/MWh, and in its optimistic case $30/MWh and $80/MWh. Thirty dollars a megawatt-hour is inside Lazard's unsubsidised utility-solar band of $38–78/MWh.

Established So the correct statement is not that NASA found space solar uneconomic. It is that NASA found the cost to be set almost entirely by assumptions for which there is no evidence, and that under its evidenced assumptions the answer is one to two orders of magnitude too expensive. A fifty-three-to-one spread within one analysis of one design is a finding about the state of knowledge, not about the hardware, and a brief that quotes only the $610 figure is as misleading as one that quotes only the $30.

Established Launch price is the whole argument, and no party is arguing from an observed price. NASA assumed $1,000/kg to low Earth orbit with a block-buy discount, and found launch to be 71% and 77% of total system cost for its two designs. The Caltech group's peer-reviewed technoeconomic paper assumed Starship at $7 million for 54,500 kg to geostationary orbit ≈ $128/kg, with an asymptotic case near $55/kg. The best price actually observed for the relevant orbit is Falcon Heavy at about $97 million for 13,000 kg to geostationary transfer ≈ $7,460/kg, and the Caltech paper's own “current” case returns $7,780/MWh. The required price is therefore between roughly seven and sixty times below the best observed price, depending on the study. Advocates dispute NASA's assumed price by substituting a different assumed price. That is not evidence, it is a preference.

Established Advocacy launch counts exclude the propellant flights. NASA's accounting is that twelve of every thirteen launches carry only fuel for orbital transfer, giving 2,321 and 3,960 total launches for its two designs. The European consultancy figure of 86–119 launches, which is what European advocacy quotes, counts payload flights only. Applying NASA's thirteen-to-one ratio to the same 2,064 tonne payload gives roughly 1,120–1,550 total flights — arithmetic this brief performed by combining two sources, and a claim neither of them makes.

Frontier The mass estimates disagree by a factor of seven, and the newest entry makes it worse. The European consultancy design implies 1,433 tonnes per gigawatt; NASA's two designs imply 2,950 and 5,000; the Caltech middle case implies 5,053. Into that spread the United States Department of Energy's own explainer describes the geostationary microwave concept as reflectors spanning up to 3 km and massing over 80,000 tonnes for multiple gigawatts, which is an order of magnitude above the European figure for the same job. The Department gives no derivation, so this brief treats it as frontier rather than established — but it widens the spread rather than narrowing it, and it comes from the energy ministry of the country that would have to build it.

Frontier Two sources inside the European programme disagree by a factor of five about the most basic engineering quantity in the system. The retired head of space power systems at the European Space Agency states that delivering 1 GW to the grid requires handling about 9 GW on board. The British design derived from the agency's own cost-benefit contractor is reported as harvesting 3.4 GW and delivering 2 GW — a ratio of 1.7 to one. Both cannot be right, and the difference is how much power the satellite must generate, switch, convert and radiate as waste heat. Neither derivation is inspectable from public documents. This brief reports both and resolves neither, and notes that it bears directly on the same author's other finding: that no electronic switches, relays or transformers have been designed or demonstrated for spacecraft anywhere near the required voltages and currents.

Established The independent modelling is not dismissive, and that is why it carries weight. A peer-reviewed European capacity-expansion study using NASA's designs across thirty-three countries finds that the heliostat-swarm design becomes complementary at about fourteen times and dominant at about nine times the 2050 terrestrial photovoltaic capital cost, and that if those thresholds were met it would cut total system cost 7–15%, displace up to 80% of intermittent wind and solar, and reduce battery deployment by over 70%. It also finds that the planar-array design never appears in a cost-optimal scenario, that long-duration storage such as hydrogen remains essential for seasonal balancing regardless, and that the technology sits one to two orders of magnitude above break-even. That is what a serious negative result looks like: it specifies the conditions under which it would become positive.

The genuinely open questions in this subject are not the ones the field advertises. What is advertised as open — can we beam power from orbit, can we build large structures in space — is answered in principle and unanswered in practice at a scale nobody has approached. What is actually open divides into four questions with different characters, and only one of them is a physics question.

Frontier The first is the conversion chain, and it is an engineering frontier with a measurable target. The 54.18% direct-current-to-direct-current figure that circulates in this literature is a 1975 NASA/Raytheon bench test conducted over metres, not a link through an atmosphere. Projections through a real geostationary link are 20–25% DC-to-DC. The full sunlight-to-grid chain, as computed by the Caltech group — the most pro-space-solar peer-reviewed team in the field, not a critic — is 2.77% current, 4.14% middle case, 6.63% asymptotic. The best measured far-field figure anywhere is Xidian University's 20.8% DC-to-DC at just over 100 m, delivering 1,180 W from a 1.2 m transmit aperture into a 5.2 m rectenna with about 87% RF-to-DC conversion at the rectenna. That geometry does not transfer to a 35,786 km link, and the result is reported by a state information office with no peer-reviewed paper identified and no independent replication located. Raising the end-to-end chain is a real frontier with real progress in it.

Frontier The second is orbital assembly at rate, and it is the frontier nobody quantifies. NASA's designs require 1.46 to 2 million modules assembled in space. China's announced 2030 geostationary mission is explicitly stated to require on-orbit assembly. The British study named robotic in-orbit assembly as one of two immature-but-essential capabilities. Assembling two million modules across a ten-year design life implies completing one every few minutes, continuously, in geostationary orbit, by machines that must survive roughly 300 °C temperature swings a couple of dozen times a year during eclipse seasons. No source consulted for this brief gives robot-hours, an assembly duration, or a maintenance concept in quantitative terms. Against that, the largest structure the field's flagship demonstration deployed was 1.8 m square and it jammed twice. The gap between the requirement and the demonstrated capability is the largest in the subject and it is the least discussed.

Frontier The third is high-voltage, high-current power handling in a spacecraft, and it may be the hardest. The retired head of space power systems at the European Space Agency states flatly that “no electronic switches, relays, and transformers have been designed or demonstrated for spacecraft that can handle voltages and currents anywhere near the required magnitude.” This is a components problem, not a systems-integration problem, and it does not get easier with scale. It is compounded by the thermal problem: whatever fraction of on-board power is not radiated as beam must be radiated as heat, and the disagreement described in the position section — 9 GW on board per grid gigawatt against 1.7 GW — is precisely a disagreement about how much heat.

Frontier The fourth is frequency, and the field has not converged. Classical reference designs use 2.45 GHz, with 5.8 GHz for smaller systems; the Caltech technoeconomic study uses 10 GHz; MAPLE flew at 9.984 GHz; Japan's undelivered demonstrator is designed for 5.8 GHz with a 2.45 GHz pilot; and a European Innovation Council project running to 2028 is developing a rectenna at 100 GHz. Frequency sets aperture, rectenna area and atmospheric loss simultaneously, so a factor of forty in frequency is a completely different machine. That the field is still exploring a fortyfold frequency range fifty-eight years after Glaser is a fair measure of how unsettled the architecture is.

Speculative Whether the rectenna site can be dual-used is open and is being investigated for the first time. A doctoral project funded by a Northern Ireland government department and beginning in October 2026 asks whether sustainable agriculture can operate beneath a rectenna, on the premise that these are fields of order 100 km² each and “the potential to use this land productively remains almost entirely unexplored.” This is a reasonable question and the honest answer today is that nobody knows, because no rectenna of any size exists. Treat any claim that rectennas are “largely transparent” and the land beneath them usable as a design intention rather than a finding.

Handwave What is not open, despite being presented as open: whether atmospheric loss is a barrier. The United States Naval Research Laboratory's 10 GHz terrestrial beaming demonstration measured that “even in heavy rainfall, loss of power is less than five percent,” and at 2.45 GHz the loss is lower still. Atmosphere is the one part of this system that behaves well. Framing it as a research frontier misdirects attention from aperture, conversion, mass, launch price and land, which are where the difficulty actually is. The 100 GHz architectures are the exception, since sub-terahertz propagation through weather is a genuinely different problem, and that work is at the rectenna-component stage.

Established The first bottleneck is launch price, and it binds because it is 71–77% of the cost. NASA's own decomposition puts launch at 71% of its heliostat design and 77% of its planar design, at an assumed $1,000/kg to low Earth orbit. Nothing else in the system has that leverage. It follows that every argument about space solar economics is, structurally, an argument about launch prices — and that the argument is currently conducted between assumed prices, since the best observed price to the relevant orbit is about $7,460/kg and every viable case assumes something between $55 and $1,000.

Established The second is mass, because mass is what launch price multiplies. Estimates per gigawatt run from 1,433 to 5,053 tonnes across four studies, with a government explainer implying an order of magnitude more. Until that spread narrows, no cost estimate in this field can be evaluated, because the two dominant terms in the cost — dollars per kilogram and kilograms — are both unknown to within roughly an order of magnitude each. Multiplying two such numbers is how a fiftyfold LCOE spread arises from one report.

Frontier The third is orbital assembly, and it binds on rate rather than on capability. One to two million modules per satellite, assembled in geostationary orbit, is a throughput requirement, and no published figure exists for the throughput of any in-space assembly system. The European response to date has been a call worth €3 million in total and a British national space agency contribution of an undisclosed amount. That is a study-level response to a capability that the same studies identify as one of two blocking dependencies.

Frontier The fourth is high-voltage power handling and the thermal management that goes with it. Components do not exist at the required ratings, according to the most senior insider critic in the field; and the required rating is itself disputed by a factor of five. This is the bottleneck most likely to be underestimated, because it is invisible in a block diagram and it does not appear in launch-cost sensitivity analyses at all.

Established The fifth is land and consent for the rectenna, which is a planning problem rather than a technical one. Thirty-four to a hundred square kilometres per satellite, inside a beam whose centre exceeds the general-public exposure limit by more than twenty times, sited near a grid connection capable of accepting a gigawatt or two. No such site has been identified, permitted or acquired anywhere, and the first serious research on whether the land can be dual-used starts in October 2026. In a corpus that repeatedly finds infrastructure blocked by siting rather than by physics, this is the bottleneck most likely to be decisive and least likely to be modelled.

Established The sixth is that the buyer does not exist. No utility, regulator or transmission operator anywhere is a counterparty to any space solar programme. The United States Department of Energy states in its own explainer that it “is not actively researching SBSP” and “isn't currently developing any SBSP technologies specifically.” The institutional section takes this up in full; it belongs in the bottleneck list because a technology whose entire justification is the electricity market has no relationship with anyone who buys electricity.

Established What is not a bottleneck, stated plainly. Atmospheric loss is not a bottleneck at classical frequencies: under 5% through heavy rain at 10 GHz. Photovoltaic performance in orbit is not a bottleneck: the 2023 demonstration found low-cost gallium arsenide performing consistently well over 240 days. Beam steering to the required precision is a solved problem in phased-array radar. Orbital mechanics is not a bottleneck; it is merely a doubling, since firm power needs two satellites rather than one. The difficulties are mass, money, assembly, components and land, in that order, and none of them is a discovery waiting to happen.

Established This brief depends first on launch economics, and the dependency is quantified. Launch is 71–77% of system cost in the only government analysis with a published decomposition. A fully reusable heavy vehicle delivering to geostationary orbit at a published, audited price is the precondition for every viable cost case, and it is a precondition this field does not control and cannot accelerate.

Established Second, on robotic in-space assembly and servicing. One to two million modules per satellite, in geostationary orbit, over years, with maintenance across a design life that NASA assumed at ten years and advocates argue should be thirty. The British study named this as one of two blocking dependencies alongside launch; the European response has so far been a €3 million call.

Established Third, on power electronics that do not exist for spacecraft. Switches, relays and transformers at the required voltages and currents, in vacuum, across 300 °C eclipse thermal cycling. This dependency is on a components industry rather than on a space programme, and it has no obvious sponsor because no other spacecraft needs it.

Established Fourth, on wireless energy transmission for the conversion chain itself. Rectenna efficiency, phased-array transmitter efficiency and beam-shaping all belong to that brief's subject matter, and the best measured far-field numbers in the world are terrestrial. Space solar cannot be more efficient end-to-end than the beaming technology it is built from, and at present that technology's best verified far-field result is 20.8% at just over a hundred metres.

Frontier What depends on it is currently very little, and that asymmetry is the finding. No grid plan anywhere counts on space solar. The European modelling that treats it seriously does so conditionally, at cost thresholds nine to fourteen times below current projections, and finds that long-duration storage remains essential even then. The clearest statement of the dependency structure is that space-based solar power waits on four other industries and nothing waits on it.

Established The decisive experiment has already been run once and it produced a negative result that the field has not fully absorbed. The 2023 Caltech mission is the only space-to-ground beaming attempt on record. It returned a detection rather than a transfer, lost 14.7% of transmitted power in eight months, and had its deployable structure jam twice at 1.8 m in ways the laboratory had not predicted. That is a rich negative result, honestly reported by the people who paid for it, and it is worth more to a reader than any roadmap.

Frontier The next real test is Japan's OHISAMA, and it is the most informative near-term experiment because it is small enough to be honest. A roughly 180 kg satellite with 2 m² of array generating about 1 kW in short bursts, in a 400 km orbit, beaming at 5.8 GHz with a 2.45 GHz pilot to thirteen receiving stations at Suwa covering 600 m² in total, with a stated beam-pointing requirement of 0.001 degrees. It has not flown. The figure that will settle the argument — how much direct current the ground stations actually recover — is not published by anyone, which is itself worth recording. If OHISAMA flies and reports a recovered-power number honestly, it will be the first genuine space-to-ground power measurement in the history of the field.

Established A natural experiment is already running in the capital markets, and it has produced a result. Total disclosed private capital across the named microwave space-solar companies is well under $100 million, against a first-plant estimate of at least $275 billion. The sector's largest raise — a reported $250–350 million Series B at a $2 billion valuation in March 2026 — belongs to a laser company that has since redirected toward orbital data centres, its founder observing that if you wanted to power artificial intelligence “it would be much more advantageous to actually put the chips in space.” When the best-funded firm in your sector concludes that it is easier to move the load to the power than the power to the load, that is a market experiment reporting a result.

Frontier The Chinese ground programme is the most substantive experimental line and the least verifiable. Xidian University's full-chain ground platform completed system testing in Xi'an in June 2022 and reported 1,180 W over more than 100 m at 20.8% DC-to-DC in May 2026, with a 1.2 m transmit array, a 5.2 m rectenna and about 87% RF-to-DC conversion, plus 143 W to a drone at 30 m. The Bishan facility in Chongqing, roughly 33 acres, is reported to have conducted a 300 m microwave transmission test in August 2021 and to include study of the effects of microwave beams on living organisms. All of this is reported through state channels; no peer-reviewed paper has been identified and no independent replication located, and whether the facility's stratospheric-platform stage was ever completed remains unverified across sources from 2022, 2025 and 2026.

Frontier The experiment that would actually settle the economics is not an experiment at all. It is a published, audited price for delivering a tonne to geostationary orbit on a fully reusable vehicle. Every viable cost case in this field depends on that number, and the single most consequential unresolved question in the research — whether the Caltech group's $7 million-per-Starship-to-geostationary figure includes tanker flights — could not be determined from public documents. If it excludes them, that paper's levelised cost is understated by roughly an order of magnitude. This brief cannot resolve it and states it as the open question it is.

Established The system is four segments and the ground segment is the one that gets omitted. A solar power satellite collects sunlight, converts it to radio frequency, radiates it through a transmitting aperture, and a rectifying antenna on the ground converts it back to direct current for the grid. Beam width is set by diffraction, so aperture and frequency together fix the size of the ground station. The published reference designs are consistent about this and public summaries are consistently silent about it.

Established Three to ten kilometres in diameter is seven to seventy-nine square kilometres, and a hundred square kilometres is the land area of a mid-sized city. Per satellite. Permanently. The two entries at the bottom of that table are the useful ones precisely because neither party is selling a satellite: one is a government energy department's public explainer, the other a university advertising funded doctoral research on the land-use problem.

Established The beam is not sunlight and the comparison to sunlight is the field's most durable rhetorical error. Reference designs put 230–300 W/m² at beam centre, and both a trade-press feature and a World Economic Forum article introduce the figure by noting it is “about a quarter of the strength of the midday sun.” The applicable limit is not solar irradiance. It is the International Commission on Non-Ionizing Radiation Protection's general-public reference level of 10 W/m² at these frequencies. At 230 W/m² the beam centre is twenty-three times the general-public limit and about 4.6 times the 50 W/m² occupational limit. The arithmetic is this brief's; the inputs are sourced. Tissue does not respond to 2.45 GHz the way it responds to visible light, which is the entire reason a separate limit exists. The rectenna is therefore a permanent exclusion zone of tens of square kilometres, not a solar farm you can walk across, and the 10 W/m² number that circulates as “the beam intensity” is the exposure limit and the rectenna-edge design target, not the beam.

Established The European Space Agency's own reference material is more aggressive still, and has not been updated since 2013. Its Advanced Concepts Team page describes a 2.45 GHz design with a theoretical transmission efficiency of 50–60% yielding “around 400 W electricity per square metre on Earth receivers, which is about two to three times the amount we could receive from the same area of terrestrial PV panel.” The agency presents the higher ground-level intensity as an advantage. That the most recent content on the page dates to 15 April 2013 is itself informative about how much institutional design work has been done since.

Established One satellite is not a power station. A peer-reviewed orbital analysis from Columbia finds that a single geostationary satellite is overhead a given rectenna about 41% of the time, and that a minimum of two satellites is required for continuous coverage. Reference designs quoting per-satellite capacity against a firm-power claim are quietly halving the space segment. This brief states, with every capacity figure, whether it is per satellite or system-level.

Frontier Availability differs sharply between architectures and it is the quantity that determines whether the concept has a point at all. NASA's heliostat-swarm design generates 99% of the year; its planar-array design generates 60%, and produces about one-fifth the output. The European modelling assigns them capacity factors of 99.7% and 60% and 2050 capital costs of €267.87/kW and €396.59/kW. A 60% capacity factor is a poor return on an orbital power station, because near-continuity is the only property that justifies going to orbit in the first place, and it is why the planar-array design never appears in a cost-optimal scenario in that model.

Frontier Thermal management is where the demonstrated hardware actually failed. MAPLE's 14.7% power decline over eight months was attributed to element failures and electrical–thermal interaction, and the team's named priority for the next iteration is thermal design. At geostationary altitude the problem is worse: eclipse seasons impose temperature swings of roughly 300 °C a couple of dozen times a year on a structure hundreds of metres across, with warping as the failure mode. No source consulted gives a radiator mass for a geostationary solar power satellite, in tonnes or in kilograms per kilowatt. The thermal problem is documented; the mass penalty for solving it is not, and it sits inside the same unresolved question as the 9 GW-versus-1.7 GW disagreement about on-board power.

The nearest neighbour by far is wireless energy transmission, and the relationship is supply rather than analogy. Every watt a solar power satellite delivers passes through a transmitter and a rectenna whose physics that brief owns. The two share sources deliberately and split them by question: the Xidian result at just over a hundred metres is a terrestrial record there and a ceiling estimate here; the exposure guidelines are a design constraint there and a land-use constraint here.

Grid-scale storage is adjacent as the incumbent. The entire value proposition of orbital solar is firm power without storage, so the relevant comparison is never solar against space solar; it is solar-plus-storage against space solar. Lazard's unsubsidised figures make that comparison concrete: utility solar at $38–78/MWh, solar-plus-storage at $50–131/MWh. A contractor to the European programme concluded in its own study that even optimistically space solar would be six times as expensive as a terrestrial solar farm with battery storage.

Heavy-lift launch is adjacent in the strong sense: it is not an input to this subject so much as the variable that determines whether the subject exists. A brief about space solar economics that does not foreground launch price is a brief about the wrong thing.

Orbital data centres are adjacent in the most interesting way, because they are the competing use of the same orbital real estate and the same photovoltaic hardware. The sector's best-funded company redirected from beaming power down to putting computation up during 2026. If the reason to go to orbit is abundant continuous sunlight, then moving the load into orbit avoids the space-to-ground link entirely — which is to say, it avoids the part of the system that has never worked.

Finally, radio spectrum management is adjacent and this brief has nothing verified to say about it. Whether any allocation, footnote or international agenda item exists for gigawatt-class space-to-Earth power beaming could not be established, and the brief declines to guess.

Established The defining institutional fact is that energy institutions do not fund this and space institutions do. The United States Department of Energy states in its own public explainer that “the Energy Department is not actively researching SBSP” and “isn't currently developing any SBSP technologies specifically.” NASA studied it and produced the most negative published cost figures in the field. The European Space Agency funded a preparatory activity that does not appear in its next subscription document. The one energy ministry that did spend money — the United Kingdom's Department for Energy Security and Net Zero — spent £6,478,969 across nine projects and has not funded a demonstrator. No utility, regulator or transmission operator anywhere is a counterparty to any of it.

Established The United Kingdom's spending is worth itemising, because it is the most legible public record in the field. Under four lots: wireless power transmission to Queen Mary University of London (£960,607), the University of Bristol (£353,398) and the Satellite Applications Catapult (£999,513); high-concentration photovoltaics to the University of Cambridge (£770,666) and MicroLink Devices UK (£449,955); energy systems engineering to Imperial College London (£295,194) and EDF Energy's UK research centre (£25,855); and mission architecture feasibility to the Satellite Applications Catapult (£424,989) and Space Solar Engineering (£1,198,802). That an electricity utility took a £25,855 study contract is the closest thing to utility engagement anywhere in this brief.

Established Europe's programme is best described as displaced rather than cancelled. SOLARIS funded cost-benefit studies by Frazer-Nash and Roland Berger in 2022, concept studies by Thales Alenia Space Italy with ENEL and by Arthur D. Little with ENGIE in 2023, and a €3 million in-space robotic assembly call in 2024, with the UK Space Agency separately funding British robotics work at an undisclosed amount. It was scoped to support a 2025 ministerial decision. It does not appear in the CM25 subscriptions. Meanwhile REMPOWER — €3,989,045.83, European Innovation Council, coordinated by Thales, October 2024 to September 2028 — is developing a 100 GHz rectenna for space solar. The money is smaller, the frequency is different, and the institutional home has moved from the space agency to the innovation council.

Established The European studies' own numbers explain why nobody committed. Frazer-Nash's central case is 54 gigawatt-class satellites producing €601 billion in benefits against €418 billion in costs, a net present value of €149–262 billion over 2022–2070. Roland Berger's range is €8.1 billion to build and €7.5 billion to operate at best, €33.4 billion and €31.1 billion at worst. A programme whose benefit-to-cost ratio in its own advocate's central case is 1.44 to one, on a half-century horizon, with a fourfold spread in build cost, is not a programme a finance ministry approves.

Frontier The advocacy institutions are real, organised and small. The Space Energy Initiative published a whitepaper in August 2026 asking the British government to co-fund a megawatt-scale geostationary demonstrator within five years on Space Solar's CASSIOPeiA platform, claiming a levelised cost of about £30/MWh falling toward £10/MWh and £1–2 billion of annual grid savings per 2 GW. Space Solar's own case page states that “every subsystem has already been demonstrated.” Both are interested parties; the ask is the verifiable fact, and the numbers are claims.

Established Private capital has already voted, and the vote is legible. Total disclosed funding across the named microwave space-solar companies is well under $100 million. One firm's Icelandic utility relationship is an explicitly non-binding letter of intent; another's assembly milestone was achieved in a laboratory at Culham, not in space. The only booked launch found anywhere in the sector belongs to a laser company with a $20 million seed and a 2028 SpaceX slot. And the sector's largest raise — a reported $250–350 million at a $2 billion valuation in March 2026 — belongs to a company that redirected during 2026 toward orbital data centres, on the reasoning that it is better to put the chips in space than to beam the power down.

Frontier The institution that does not exist is a regulator for the ground station. No jurisdiction has a permitting regime for a 34–100 km² receiving site inside a beam exceeding general-public exposure limits by more than twentyfold, and no such site has been identified anywhere. Nor could this brief verify that any international spectrum allocation exists for gigawatt-class space-to-Earth power beaming; that question was left open by the research and the brief declines to fill it. When a technology's most-cited obstacles are launch price and assembly, and its unexamined obstacle is a planning application nobody has ever filed, the unexamined one deserves attention.

The evidence base in this field is unusually dominated by interested parties, and the brief marks them. Established The primary flight data comes from the developers. The most quoted mass and cost figures come from a consultancy commissioned by a space agency, whose space business lead subsequently co-chairs the national advocacy body that lobbies for the technology, and which also wrote the national feasibility study. That is not misconduct; it is a small field. But it means the same analytical voice appears as agency contractor, national study author and advocate, and a reader who encounters those three documents separately will mistake one opinion for three.

Established The clearest counterweight is that the most negative published figures come from government analysts, and the most careful positive figures come from the developers. NASA's office of technology policy produced the $610 and $1,590 per megawatt-hour baselines. A retired agency power-systems chief produced the $275 billion first-plant estimate and the observation that the required power components do not exist. And the Caltech group, the most committed advocates in the peer-reviewed literature, published a sunlight-to-grid efficiency of 2.77% for the current case. Where the advocates publish an unflattering number about their own system, believe it.

Established Public money in this field is small, traceable and worth stating exactly. The United Kingdom's competition awarded £6,478,969 across nine projects. The European Innovation Council put €3,989,045.83 into a rectenna project. The European Space Agency's in-space assembly call was €3 million in total. Against a first-plant estimate of at least $275 billion, the United Kingdom's entire identified research spend is a ratio of about one to forty thousand. This brief takes no position on whether that is too much or too little; it observes that the sums are small enough that the argument about them is symbolic, and that symbolic arguments are where public deliberation goes wrong most often.

Frontier The opportunity-cost question is the substantive ethical one and it is not settled. Every pound spent on orbital solar is a pound not spent on storage, transmission, demand flexibility or terrestrial generation, all of which have shorter paths to deployment. The strongest version of the counter-argument is that the peer-reviewed modelling identifies a genuine gap — firm zero-carbon power — that storage fills expensively and that nothing else fills well. The strongest version of the sceptical case is a European contractor's own conclusion that even optimistically this would cost six times a terrestrial solar farm with batteries. Both are honestly held.

What this brief could not establish, stated as an obligation rather than a caveat. Frontier It does not know whether the required launch price will ever be charged, because no audited price for reusable delivery to geostationary orbit exists. It does not know the mass of a gigawatt-class system to better than a factor of seven, or the on-board power requirement to better than a factor of five. It does not know the radiator mass, because no source gives one. It does not know the assembly rate, because nobody has published one. It does not know whether any spectrum allocation exists for space-to-Earth power beaming, because that could not be verified and this brief will not guess. And it does not know how much power Japan's demonstrator expects to recover on the ground, which is the number its whole mission will be judged on and which nobody publishes. Six unknowns of that magnitude, in a subject with a fifty-eight-year literature, is the finding.

Established The civilisational significance of space-based solar power is not the electricity. It is that this is the cleanest available case study in how a field's evidence base can be almost entirely composed of design assumptions. Nothing here is fraudulent. The studies are competent, the engineers are serious, the physics is sound. And yet the mass estimate varies sevenfold, the on-board power requirement varies fivefold between two sources inside the same agency's orbit, the levelised cost varies fifty-three-fold inside a single report, and the only flight measurement of the core function was a detection. A reader who learns to ask “is this demonstrated, designed, or assumed?” from this subject will carry the habit usefully into a dozen others.

Frontier If it worked, the prize is specific and has been quantified by people with no stake in it. Firm solar power without storage would cut total European system cost by 7–15%, displace up to 80% of intermittent renewables and reduce battery deployment by over 70%, in a peer-reviewed thirty-three-country model — conditional on a cost reduction of one to two orders of magnitude, and with seasonal storage still required. That is a real prize and the conditions are real conditions. Both halves belong in any honest account.

Speculative There is a strategic dimension that the economics obscure. A geostationary structure that delivers gigawatts to a fixed ground site is also a piece of national infrastructure that cannot be embargoed, blockaded or easily seized, sited above a country that owns the receiver. Chinese reporting has associated renewed official interest with potential military applications, and the same aperture that delivers power can, in principle, deliver it somewhere else. This brief has no verified sourcing on the dual-use question and states it as a consideration rather than a finding.

Established The general principle this case illustrates, stated plainly. When a technology's cost is dominated by a single input price that the technology does not control — here, launch, at 71–77% of system cost — then work on the technology is not what determines whether it happens. Fifty-eight years of solar power satellite research have produced excellent engineering and no change in the variable that matters. The lesson is not that the research was wasted. It is that identifying which variable governs, and whether you can move it, is prior to deciding what to work on.

Established What has actually happened. Peter Glaser proposed the concept in 1968. A NASA/Raytheon bench demonstration in 1975 produced the 54.18% DC-to-DC figure still quoted today, at a range of metres. Japan has worked continuously since the 1980s, giving it the field's longest research record. China approved proposals in 2013 and a roadmap in 2014, began building at Bishan around 2019, and reported a 300 m ground transmission in August 2021. The European Space Agency approved SOLARIS as a preparatory activity in November 2022. Caltech flew SSPD-1 from January to November 2023. NASA published its analysis in January 2024. The United Kingdom's energy department ran a £6,478,969 innovation competition across nine projects, last updated April 2024. The European Innovation Council funded REMPOWER, a €3,989,045.83 sub-terahertz rectenna project coordinated by Thales, from October 2024 to September 2028.

Established What happened at the European ministerial, stated precisely because it is easy to get wrong. SOLARIS was scoped as a three-year study to support a decision at the 2025 ministerial. SOLARIS is absent from ESA/C-M(2025)100 REV.5, the official CM25 optional-programme subscription document of 1 December 2025, which subscribed €22.07 billion in total — no mention of SOLARIS, space-based solar power, solar power satellites or power beaming anywhere in it. The agency's SOLARIS page carries no dated item after July 2023. This brief does not say CM25 funded it and does not say it was cancelled, because no statement to either effect exists; activities could in principle continue unnamed inside the general technology programme, and that is unverifiable either way.

Handwave Japan, 2026. OHISAMA has slipped from 2025 to 2026 and 2026 coverage still describes it as planned. No launch vehicle has been identified in any source consulted. Flagged handwave as a date, established as a design.

Handwave China: 2028 low Earth orbit at 10 kW over 400 km; 2030 geostationary at up to 1 MW requiring on-orbit assembly; 2035 at 10 MW with arrays over 100 m; 2050 at 2 GW with arrays around 1 km. Reported from the China Academy of Space Technology, with the reporting trade publication's own caveat attached: the project is “far from certain to proceed or gain official approval.” The separate “Three Gorges Dam in space” framing traces to a single October 2024 lecture remark and has no budget, timeline or formal approval attached to it.

Handwave United Kingdom, five years to a megawatt-scale geostationary demonstrator. In August 2026 the Space Energy Initiative published a whitepaper asking the government to co-fund exactly that, on Space Solar's CASSIOPeiA platform. The earlier public case, from 2021, proposed deployment within eighteen years — that is, by about 2039 — on a £350 million public Phase 1 over five years. Neither the Phase 1 nor the demonstrator has been funded. Flagged handwave because the schedule has already been restated once with the clock reset.

Speculative 2030s and beyond. A gigawatt-class operating system requires, in series, a launch price roughly an order of magnitude below anything observed, a mass estimate that stops varying sevenfold, in-space assembly at a rate nobody has published, spacecraft power components that do not exist, and a permitted ground site of tens of square kilometres. Each step is individually conceivable; the conjunction on a decadal timescale is speculative, and no source consulted proposes a critical path with dates attached to all five.

• Depends on Wireless energy transmission, without qualification: the rectenna, the phased-array transmitter and the RF conversion chain are that brief's subject, and this one inherits their efficiency ceiling. The relationship is unusually clean, because the two briefs split the same physics by altitude rather than by discipline.
• Requires (not on this map) A fully reusable heavy launch vehicle delivering to geostationary orbit at a published price, since launch is 71–77% of system cost and every viable case assumes a price nobody has been charged. Robotic in-space assembly at a throughput of one to two million modules per satellite, for which no rate figure has ever been published. Spacecraft power electronics — switches, relays, transformers — at voltages and currents for which, on the most senior insider account available, nothing has been designed or demonstrated. A transparent launch market, because the difference between $55/kg and $7,460/kg is the difference between this working and not. And a siting and consent regime for a ground station of 34–100 km² whose beam centre exceeds the general-public exposure limit twentyfold, which no jurisdiction has. All five are industrial, financial or institutional, and none is a discovery.
• Enables Firm, dispatchable solar power without storage — the one property that would justify the whole enterprise. The peer-reviewed European modelling puts a number on the prize: 7–15% total system cost reduction, up to 80% displacement of intermittent wind and solar, over 70% less battery deployment — conditional on capital costs falling to nine to fourteen times 2050 terrestrial photovoltaic, and with long-duration seasonal storage still required. No typed enabling edge is claimed, because the enabled endpoint is a grid outcome rather than a brief.
• Adjacent Wireless energy transmission, the sibling slot and the supplier of the conversion chain; grid-scale storage, which is the incumbent solution to the same problem and the benchmark space solar must beat; heavy-lift launch, which is the actual determinant of the economics; and orbital data centres, which are the same industry's competing answer to the question of what to do with power in space — and which, on the evidence of where the sector's capital went in 2026, is winning the argument.

“Caltech beamed solar power from space to Earth in 2023.” Established It beamed a detectable signal. The primary paper defines power transfer as “a net positive recovery” and detection as “a process where the recovered power is less than the external power needed for the receiver,” and classifies the Earth experiment as the latter. Caltech's own release said “detectable power.” The coverage that turned this into solar power collected by photovoltaic cells and transmitted to Earth by microwave beam is contradicted by the source it is reporting on. The in-space link was real and measured — 251 mW at about 30 cm — and it is the ground result that has been misdescribed.

“MAPLE proved the physics of the space-to-ground link.” Established It exercised a link at 9.984 GHz from a 160 × 90 mm array, and the ground station was never inside the beam's −3 dB contour because the host spacecraft could not hold alignment. Classical reference designs use 2.45 or 5.8 GHz with apertures of 750 m to 2.4 km. Frequency and aperture jointly determine beam width, rectenna area and atmospheric loss, so the demonstration did not test the configuration any operational system would use.

“NASA found that space-based solar power is uneconomic.” Established This is half of NASA's finding and the half that is missing is the more interesting one. The baselines are $610 and $1,590 per megawatt-hour, which NASA described as 12–31 and 32–80 times higher than 2050 terrestrial projections. The same report's sensitivity analysis, with electric propulsion, lower launch cost and longer component life, reaches $200 and $500 per megawatt-hour, and in its optimistic case $30 and $80 — the first of which sits inside Lazard's unsubsidised utility-solar band. A fifty-three-to-one spread inside a single analysis of a single design is a statement about how little is known, not about how bad the technology is.

“Launch costs are about to fall far enough to make this work.” Handwave Possibly, but nobody in this argument is reasoning from an observed price. NASA assumed $1,000/kg to low Earth orbit; the Caltech technoeconomic paper assumed $128/kg to geostationary orbit and an asymptotic $55/kg; the best observed price to the relevant orbit is about $7,460/kg. Critics of NASA's pessimism dispute an assumed number by substituting another assumed number. And the single most consequential unresolved question in the research for this brief is whether the $7 million-per-flight figure behind the $128/kg includes propellant tanker flights — because if it does not, that paper's levelised cost is understated by roughly an order of magnitude.

“It only takes about a hundred launches.” Established That count is payload flights only. NASA's accounting, which includes orbital-transfer propellant, is that twelve of every thirteen launches carry only fuel, giving 2,321 and 3,960 total launches for its two designs. Applying the same ratio to the European design's payload mass gives roughly 1,120–1,550 flights — this brief's arithmetic from two sourced inputs, and a claim neither source makes.

“The beam is only a quarter as intense as midday sunlight, so it is safe.” Established The comparison substitutes the wrong limit and it is the most durable error in this literature. At 230 W/m² the beam centre is twenty-three times the ICNIRP general-public reference level of 10 W/m² and roughly 4.6 times the 50 W/m² occupational limit. Sunlight and 2.45 GHz microwaves interact with tissue by different mechanisms, which is why they have different limits. The 10 W/m² figure that circulates as “the beam intensity” is the exposure limit and the rectenna-edge design target, not the beam.

“The ground station is compact and the land can be shared.” Established The published footprints are 4 km in diameter at 2.45 GHz, over 34 km² in a Thales Alenia design, about 50 km² in the European consultancy study, 3–10 km in diameter in the US Department of Energy's own explainer, and 10 km × 10 km in a British university's statement of the problem. Whether anything useful can happen underneath is an open research question that a funded doctoral project begins investigating in October 2026, on the explicit premise that it “remains almost entirely unexplored.” Claims that rectennas are largely transparent and the land beneath them productive are design intentions, not findings.

“Space solar could cut Europe's renewable energy needs by 80%.” Established The study behind that sentence says something considerably narrower. In a thirty-three-country capacity-expansion model, NASA's heliostat design displaces up to 80% of intermittent wind and solar — conditional on its capital cost falling to nine to fourteen times 2050 terrestrial photovoltaic cost, which the same paper says is one to two orders of magnitude away — and long-duration storage such as hydrogen remains essential for seasonal balancing even then. The other reference design never enters a cost-optimal solution at all.

“Every subsystem has already been demonstrated; this is an engineering scale-up.” Handwave This is a developer's claim and it does not survive contact with the flight record. The deployable-structure subsystem's flight demonstration was 1.8 m square and jammed twice in ways laboratory testing had not produced. The space-to-ground beaming subsystem's flight demonstration was a detection. The high-voltage power-handling subsystem is the one a retired agency power-systems chief says has never been designed for spacecraft at anything near the required magnitude. Demonstrating a subsystem at one part in a thousand of its required scale is not the same as demonstrating it.

“Europe cancelled space solar” — or, equally, “Europe funded it at the 2025 ministerial.” Established Neither statement exists in any document. What is verifiable: SOLARIS does not appear anywhere in the CM25 subscription document of 1 December 2025; the agency's SOLARIS page has no dated item after July 2023; and a separate European Innovation Council project, REMPOWER, is running to September 2028 with €3,989,045.83 of European Union money under Thales's coordination, developing a 100 GHz rectenna. European work continues under a different name and at a frequency forty times the classical one.

“China is building a Three Gorges Dam in space.” Handwave That phrase traces to a single lecture remark in October 2024, comparing the scale to “moving the Three Gorges Dam to geostationary orbit 36,000 km above the Earth.” No budget, timeline or formal approval attaches to it. The programme that does exist is a roadmap — 10 kW in low Earth orbit in 2028, up to 1 MW in geostationary orbit in 2030, 10 MW in 2035, 2 GW in 2050 — which the trade publication reporting it describes as “far from certain to proceed or gain official approval.” A related name, “Long Xiang,” could not be verified as a Chinese space solar project at all and this brief does not use it.

“Japan has beamed solar power from space.” Established It has not. OHISAMA has not launched. The story that circulates conflates a December 2024 aircraft-to-ground test from roughly 7 km altitude, which is a different experiment on a different scale in a different medium.

“The atmosphere is the hard part.” Established It is the easy part. A naval research laboratory's 10 GHz beaming demonstration measured under 5% loss through heavy rainfall, and 2.45 GHz does better. The hard parts are mass, launch price, assembly rate, power components and land.