On 21 September 2026 Katalyst Space announced that the US Department of Defense had selected it to

demonstrate space-based power beaming. The government's own solicitation for that programme is

public, and it says something narrower: the primary focus is the platform, and power beaming is the

payload that proves it.

Figures below are labelled documented where a source states them and estimated where I have

calculated them from stated assumptions. Nothing here is a measurement.

What the government asked for, in its own words

Katalyst's release of 21 September 2026 says the company will build and launch a robotic spacecraft

that will "assemble a large-format power beaming system on orbit and transmit power back to Earth",

under a programme called ASGARD, funded by the Operational Energy Capability Improvement Fund, using

modular tiles built by the US Naval Research Laboratory. It publishes no power level, no frequency,

no altitude, no tile count, no contract value and no demonstration year. RuthAnne Darling, the

Director of Operational Energy - Innovation, is quoted in it saying the work "opens new pathways for

operational energy in space".

The office she directs published the solicitation. Request for Project Proposal HQ0034269ASGD,

issued by Washington Headquarters Services on behalf of that office and still linked from the fund's

own portal at oecif.org, carries the same expansion of the acronym, Advanced Scalable Growth

Architecture for Robotic Deployment, and describes the same LEO spacecraft assembling

government-furnished modular panels with robotic arms. It sets a white paper deadline of 24 June

2026 and anticipates a prototype Other Transaction agreement under 10 U.S.C. § 4022.

Its framing is the interesting part. The document says the platform could host applications such as

on-orbit solar power collection and beaming, and then states that "the primary focus of the ASGARD

program is the demonstration of the foundational persistent platform itself". Power beaming appears

as the second of three key objectives, under the heading of demonstrating application-specific

operation as a testbed. The tiles are not the contractor's to design: the solicitation says the

"high efficiency Solar to RF tiles to beam power from LEO to the ground" will be government

furnished and integrated into government-furnished panels.

None of that makes the award less real, and it is not a discrepancy with the company's release.

Assembling a large structure in orbit is the harder and more general problem, and it is the one the

government wrote down first. But it changes what a reader should expect to see at the end. A

programme whose primary objective is a persistent modular platform succeeds if panels get joined,

powered and pointed. It does not need a single watt to arrive anywhere useful on the ground.

The fund has been buying in this direction deliberately. Its FY26 call for proposals, dated 7 May

2025, lists under the emerging-battlefields priority: "Space - advanced space qualified photovoltaic

(PV); in-space recharging/refueling; power beaming". That call caps a multi-year OECIF contribution

at $3m a year for up to three years, or $5m for a one-year surge, though ASGARD arrived through a

targeted solicitation rather than the annual cycle, so whether those caps bind it is not established

here. The novelty is not that the Pentagon is interested in beamed power. It is the robot.

The one physical dimension in the document decides the rest

The solicitation contains exactly one hard physical number, and it is a good one. The robotic arms

must "manipulate government furnished modular panels with a maximum dimension of 0.7m x 0.9m x

0.2m". That is documented. Everything from here is estimated, and the assumptions are stated as they

are used.

A 0.7 by 0.9 metre panel presents at most 0.63 square metres of aperture. A hundred of them, joined

edge to edge with no loss, make 63 square metres, which is a circle about 9 metres across. Assume a

500 kilometre orbit, which is the middle of LEO and the solicitation names no altitude. At that

range a 9 metre aperture is comfortably in the far field, so the ordinary expressions apply.

The width of the beam at the ground follows from the aperture and the wavelength, not from the

power. For a uniformly illuminated circular aperture the first null sits at about 2.44 λR/D. The

solicitation names no frequency, so take the two that the field actually uses. At 2.45 GHz that

hundred-panel array puts a spot about 16.7 kilometres across on the ground. At 10 GHz it is about

4.1 kilometres.

Now the power density, which is what a receiving station cares about. On axis it is roughly

P·A/(λR)², where A is the transmitting aperture. Radiate a kilowatt from those 63 square metres at

2.45 GHz from 500 kilometres and the peak at the ground is about 17 microwatts per square metre. A

generous 100 square metre rectenna sitting exactly under it collects about 1.7 milliwatts. At 10 GHz

the same array gives about 0.28 milliwatts per square metre and about 28 milliwatts captured. These

are estimates, and they are dominated by the aperture, so an error in my assumed efficiency does not

move them. An error in the assumed frequency moves them by the square of the ratio.

Run it backwards and the scale of the ambition becomes legible. To put a 100 metre spot on the

ground from 500 kilometres you need a transmitting aperture of about 366 metres at 10 GHz, or about

1.5 kilometres at 2.45 GHz: roughly 167,000 panels of 0.63 square metres for the first, and 2.8

million for the second. That is why the programme is about robotic assembly rather than about

transmitters. Nobody launches that aperture. ASGARD asks whether a machine can build one a panel at

a time, and nobody has yet built anything at all that way.

There is a second requirement in the document that tells you where the risk sits. The spacecraft

must "incorporate a sensor on the bus to measure planarity of the payload's antenna (non-PV)

surface". Planarity is a proxy for phase. If the panels are to combine coherently, their placement

error has to be a small fraction of a wavelength. Take a sixteenth of a wavelength as a working

tolerance, which is estimated and on the loose side: that is about 7.6 millimetres at 2.45 GHz and

about 1.9 millimetres at 10 GHz. A robot arm placing panels across tens of metres to millimetre

flatness is the actual experiment. The beam is the instrument that reads out whether it worked.

A band nobody has named, and no allocation to name it in

Neither the company nor the solicitation names a frequency, and there is a reason it is easier to

leave blank. Wireless power transmission has no status in the Radio Regulations. The current text is

the Edition of 2024, which the ITU describes as the Regulations adopted by WRC-95 and reviewed by

the conferences through WRC-23, and it carries no allocation for beaming power from space to Earth.

The band most often assumed for this work, 5.8 GHz, is not regulatorily available for space-to-Earth

transmission, and saying so is not a technicality: it is the difference between an experiment under

a national licence and a service.

The only ITU-R Recommendation naming frequency ranges for beam wireless power transmission is

SM.2151-0 of September 2022, listed as In force when I checked the editions page on 26 September

2026, and its title scopes it to "mobile/portable devices and sensor networks". A solar-to-RF tile

array in low Earth orbit is not that. The broader question is queued behind a conference: WRC-31

preliminary agenda item 2.2, under Resolution 910 (WRC-23) and assigned to Working Party 1A, is

framed as considering bands for non-beam and beam wireless power transmission "to avoid harmful

interference to the radiocommunication services". That is a protection item, not an enabling one,

and there is no space-based solar power item beside it.

Domestically the record is just as empty. A full-text search of the Federal Register for "power

beaming", across every agency, for anything published since 1 January 2026, returned zero documents

when I ran it on 26 September 2026. The venue for a demonstration like this is an experimental

licence, and an experimental licence is not a rulemaking. So the first real space-to-Earth power

beam over US soil may well be authorised without any public rule ever being written. Whoever

publishes the conditions attached to that licence, with the band and the permitted power flux

density in them, will have done the field a larger favour than the demonstration itself.

What this demonstration can honestly prove

Two constraints decide what the phrase "power to remote locations" can mean here, and neither is

efficiency.

The first is exposure, and it is the constraint that turns out not to bind. ICNIRP's 2020

guidelines set, in Table 5, a whole-body incident power density reference level of 10 watts per

square metre for the general public and 50 for occupational exposure, between 2 and 300 GHz,

averaged over 30 minutes. IEEE C95.1-2019 sits alongside them, still an active standard, board

approved in February 2019 and published in October 2019. Put the estimate above beside the limit:

17 microwatts per square metre is roughly 600,000 times below the general public reference level.

Nothing about a hundred-panel array in LEO is an exposure problem. The exposure problem arrives with

the kilometre-scale aperture, not with this one, and it will be a genuine one then.

The second is geometry, and it binds hard. A spacecraft in a 500 kilometre circular orbit has a

period of about 94.5 minutes and is above the horizon of a fixed ground site for at most about 11.6

minutes, on the single pass that goes through the zenith. Most passes are shorter and most are not

overhead. That is estimated from circular-orbit geometry alone, not from any access model of ours.

A single spacecraft therefore delivers minutes per day to any one place, separated by hours, and the

number that matters operationally is the typical worst gap in a day rather than any average. Beamed

power from one LEO platform is not a substitute for a generator at a forward site. It is a

demonstration of a link, on a schedule someone has to plan around, which is the same scheduling

problem every other orbital service has.

The case against everything I have written here is worth stating plainly, because it is decent. If

the objective is to prove robotic assembly and coherent operation, then the spot size at the ground

is beside the point, and criticising the delivered milliwatts is like criticising the first

transatlantic cable for its bit rate. That is fair. It is also exactly why the framing in the

solicitation deserves to travel further than the framing in the press release. The document says the

platform is the product. Read that way, the programme is well posed. Read the other way, as the

Pentagon buying beamed power for remote sites, it invites a disappointment it did not earn.

BeamDesk has no involvement in this programme and no relationship with any organisation named here.

Everything above comes from published material and from the issuing bodies' own documents. One loose

end, stated as loose: the solicitation does not identify the tile design, and the US Navy's own

DVIDS entry of 18 May 2020 records its Photovoltaic Radio-frequency Antenna Module being deployed

aboard an X-37B, with the aim of a later spacecraft "to test the transmission of energy back to

Earth". Whether ASGARD's tiles descend from that module is not stated in any source I have seen,

and I am not going to assume it.

What would change my mind

If the government or the contractor publishes a target power flux density at the ground, with the

band and the assembled aperture area, the second section becomes checkable rather than indicative,

and if those figures imply a spot far smaller than my estimate then the assumption that broke is

probably the panel count or the frequency, and I would want to know which.

If the demonstration turns out to be aimed at a receiving site with a rectenna measured in hectares

rather than in hundreds of square metres, the captured-power figures move by orders of magnitude and

the "milliwatts" framing here is too harsh.

If a follow-on production transaction under 10 U.S.C. § 4022(f) is announced with a power delivery

requirement in it, then the platform-first reading of this programme was correct only for the

prototype, and the fund is buying beamed power after all.

If the FCC grants an experimental licence for a space-to-Earth power beam and publishes the

conditions, or if Working Party 1A receives a contribution proposing a space-to-Earth band rather

than an interference-protection study, then the regulatory vacuum described above has started to

close, and that is the single most consequential thing that could happen in this field.

Sources