On 30 September Virtus Solis and Brae Systems announced a twenty-year power purchase
\nagreement for 100 MW of orbital solar power delivered to a receiving station in Illinois.
\nThe release's own annual energy figure, 876,000 MWh, is 100 MW multiplied by all 8,760
\nhours of a year.
\nThat is a capacity factor of exactly one. The expansion option in the same release carries
\nthe same property: 250 MW and 2.19 million MWh, which is 250 multiplied by 8,760 to the
\ndigit. Both figures are documented, in the sense that they are published by the parties
\nthemselves. Neither is a measurement, and neither is an estimate of what will arrive at a
\nrectenna in Illinois.
\nThe release also states a 99.9% uptime target. Those two numbers do not describe the same
\nyear. Take 99.9% of 8,760 hours and you get 8,751.2, which at 100 MW is 875,124 MWh. The
\nannual figure the release publishes is 876 MWh higher, because it was not derived from the
\navailability target at all. It was derived from the nameplate.
\nThe gap is small and the point is not that 876 MWh matters. The point is the method. A
\nterrestrial contract states contracted capacity, then an availability obligation, then an
\nexpected annual delivery that is strictly less than the product of the first two, because
\nsomebody has subtracted scheduled maintenance, forced outage and conversion loss. Here the
\nsubtraction step is absent, and the word used for the result is "delivering".
\nThere are specific reasons an orbital number cannot be 8,760 hours. According to pv
\nmagazine USA, reporting on 1 October 2026, the architecture is a Molniya-type highly
\nelliptical orbit with a twelve-hour period, each orbital power station visible from the
\nground site for about eight of those twelve hours, and three stations required for a
\ncontinuous supply. BeamDesk has not independently verified that description against a
\nfirst-party Virtus Solis publication, because the company's own news index has not been
\nupdated since 2024. If it is right, three stations at eight hours each is twenty-four
\nhours with zero margin: the architecture as described has no spare satellite, and any
\nstation out of service is a gap in a product sold as baseload. That is before eclipse
\nseasons, before rain and atmospheric absorption on the downlink, before handover between
\nstations, and before DC conversion at the receiver.
\nNone of this makes the contract dishonest. Rated power times hours is a perfectly normal
\nway to express a ceiling. It is the wrong way to express a delivery, and a reader who
\ntakes 876,000 MWh as a forecast has been handed the single most misleading collapse
\navailable in this field: rated quoted as though it were average.
\nThe coverage calls this the first commercial orbital solar power purchase agreement. It is
\nnot, and the earlier one is worth reading because it did the arithmetic better.
\nOn 3 December 2009 the California Public Utilities Commission issued Resolution E-4286,
\napproving cost recovery for a power purchase agreement between Pacific Gas and Electric and
\nSolaren Corporation. The terms in the resolution are 200 MW, a fifteen-year contract, 1,700
\nGWh per year, and a commercial operation date of June 2016.
\nTwo hundred megawatts across 8,760 hours is 1,752 GWh. The 2009 contract wrote 1,700. The
\ndiscount is three per cent, which is far too small to represent a real availability model
\nfor a beamed link, but it is not zero, and it tells you somebody in that transaction
\ndistinguished a ceiling from a delivery.
\nThe Commission also wrote down what it was approving. The resolution describes a
\n"first-of-its-kind space-based solar project" and says the Commission had previously
\napproved renewable contracts only for "projects that utilize commercialized or
\npre-commercialized technology, not technologies that are still in the research and
\ndevelopment stage". It conditioned approval accordingly: PG&E was barred from relying on
\nthe Solaren project for procurement planning until construction actually began.
\nSolaren's own account of what happened next, published on its website on 29 January 2021,
\nis that it could not obtain commercial financing, and that "In 2015, Solaren and PG&E
\nmutually agreed to forgo the power contract". No power was ever delivered.
\nThe useful difference between 2009 and 2026 is not optimism. It is that the 2009 deal
\npassed through a regulator who insisted in writing that an unbuilt orbital project could
\nnot be counted on, and the 2026 deal is between two private companies with no such referee.
\nLatitude Media reports that the agreement carries technical and financial milestones on both
\nsides and that Virtus Solis's founder expects first power around 2030. BeamDesk has not
\nindependently verified either statement; neither appears in the companies' release.
\nA power purchase agreement settles on measured delivered energy, so it is worth being
\nprecise about which part of that is actually missing for a beamed link.
\nThe meter is not missing. Direct current out of a rectenna is measurable with ordinary
\nrevenue-grade instruments, the same as any inverter output. What is missing is everything
\nthe contract has to say around it: where the delivery point sits, how availability is
\ndefined when the transmitter is in orbit and the receiver is not, and who carries a
\nshortfall when the beam, the weather and the receiver are three separate failure modes with
\nthree separate owners.
\nThere is no published method for that apportionment. The ITU-R Recommendation covering this
\nclass of system, SM.2151-0, is still the only edition in force on the ITU's own editions
\npage, and its title scopes it to "wireless power transmission via radio frequency beam for
\nmobile/portable devices and sensor networks". A hundred-megawatt orbital link is not the
\nthing that document was written about. On the measurement side, IEEE P63480, titled
\n"Standard for Assessment of Human Exposure to Electromagnetic Fields from Radiative Wireless
\nPower Transfer Systems", has been an Active PAR since 29 June 2023 with no published draft,
\nand in any case it addresses exposure assessment rather than revenue metering.
\nExposure is where the published numbers do exist, and they bind on power density rather
\nthan on contracted megawatts. ICNIRP's 2020 radiofrequency guidelines set a general public
\nreference level of 10 W/m² incident power density above 2 GHz, averaged over 30 minutes and
\nthe whole body, with a 40 W/m² local limit over 6 minutes. IEEE C95.1-2019 is the parallel
\ninstrument and remains an active standard. Neither is a constraint on how many megawatts a
\ncontract names. Both are constraints on the beam's footprint, and no spot size or edge power
\ndensity for a 100 MW link has been published by anyone in this story.
\nThe hardware step beneath the contract is documented and it is small. The US Department of
\nEnergy's NEPA categorical exclusion record CX-270832, dated 16 September 2025, describes the
\nVirtus Solis wireless power transfer project as a four-metre phased array transmitting over
\none kilowatt of radio frequency power across 200 metres to a four-metre receiving array, at
\n"minimum 70 percent end-to-end efficiency from source to delivered direct current". The
\nSeptember release points at March 2027 for "a public demonstration in which it expects to
\nyield record Wireless Power Transfer (WPT) efficiencies", and publishes no figure for them.
\nSet that beside the contract. One kilowatt becomes one hundred thousand kilowatts. Two
\nhundred metres becomes the slant range to a highly elliptical orbit: a twelve-hour period
\nimplies a semi-major axis near 26,600 km, and on a Molniya-like eccentricity that puts
\napogee in the region of 40,000 km altitude. That last figure is an estimate derived from the
\nstated period, not a published Virtus Solis number. pv magazine USA puts an intermediate
\nstep in between, a 100 kW pilot in low Earth orbit in roughly twenty-four months.
\nSo the published ladder is a kilowatt on the ground in 2027, a hundred kilowatts in orbit
\naround 2028, and a hundred megawatts around 2030. Each rung is three orders of magnitude or
\na change of regime, and only the first has a government record attached to it.
\nA milestone-conditional PPA for unbuilt capacity is a real instrument and not a press
\nrelease with a number in it. It gives a developer a named counterparty, which is what
\nproject finance asks for first, and it tells you a buyer with a specific load profile has
\ndone enough diligence to sign something. Buyers of continuous power for compute have a
\ngenuine reason to look at orbit, because the thing they are short of is firm supply rather
\nthan nameplate.
\nThe date worth putting in a calendar is March 2027. A measured end-to-end efficiency at 200
\nmetres, published with its method, would be the first number in this entire story that is
\nnot a projection. Until then the honest description of the agreement is that two private
\ncompanies have agreed terms for electricity that nobody has yet generated, metered or
\ndelivered, with an annual energy figure computed as though the sun never set and nothing
\never broke.
\nPublication of the contract's availability structure. If 876,000 MWh is a stated ceiling
\nwith an availability schedule and a shortfall remedy underneath it, then the release is
\nloose and the agreement is sound, and the criticism above lands on the press release rather
\nthan on the deal. One term sheet summary defining the delivery point would settle it.
\nA first-party availability model from Virtus Solis: eclipse seasons, handover between
\nstations, atmospheric loss at the Illinois site, with an annual energy estimate that is
\nvisibly below rated power times 8,760 hours. That would show the subtraction was done and
\nsimply not published.
\nA measured DC-to-DC efficiency at or above 70% across 200 metres at the March 2027
\ndemonstration, with the method stated. That makes the kilowatt rung real, and it would move
\nmy scepticism from the arithmetic to the much more interesting question of whether the
\nscaling holds.
\nWhat would not change my mind is more coverage of the same release, or a larger contract
\nannounced before a measured figure exists.
\n