According to Anduril, the first calibrated images have arrived from its first satellite, which carries three payloads – one long-wave infrared and two electro-optical. The company says it paid for the satellite itself, and the spacecraft reached orbit in July. Two months later, the US Space Force announced the first six operational agreements under its Commercial Augmentation Space Reserve (CASR), a pool of private companies meant to provide a commercial backup for tracking objects in orbit. One report says the firms include Anduril, EOS Space Systems, NorthStar Earth and Space, InTrack Radar, Slingshot Aerospace and LeoLabs, with the contracts worth $1.37 million in total.
The two announcements point to a new kind of arrangement. A company builds a defense-grade sensor with its own money, and the government enrolls it in a wartime reserve. Anduril was also awarded a $25 million contract in 2024 to build a mesh network for the Space Surveillance Network, which means the same firm is now a supplier, a satellite owner and a reserve member at once.
However, many in the defense community fail to recognize that such companies are not just contractors, but rather potential military targets. The agreements signed so far do not say who carries that risk.
‘A backup that may not survive’
Early CASR agreements did not cover compensation for satellites lost in war, nor did they settle whether a provider can be required to deny service to its other customers. Coverage of this autumn’s awards says those legal questions remain open. A company can therefore join the reserve without knowing who pays if its satellite is destroyed.
‘Purpose, not paperwork’
Under the law of armed conflict, an object becomes a military objective when its nature, location, purpose or use makes an effective contribution to military action and its destruction offers a definite military advantage. Scholars at the Lieber Institute and Stanford have applied this test to commercial satellites supporting Ukraine, and Russia has said something similar at the United Nations. A contract plays no part in the test. Purpose does, and a satellite designed for military sensing from the start may find it difficult to claim a purely civilian status.
‘From orbit to the ground’
The same logic has already reached the ground. In March, drones struck two Amazon data centers in the United Arab Emirates and damaged a third in Bahrain. Analysts at Georgia Tech described it as the first time a country has deliberately targeted commercial data centers in wartime. Most of the advice that followed was aimed at customers, who were told to diversify their regions and review force majeure clauses. Few asked who should carry the risk in the first place.
Data centers are a harder case than satellites. Satellites orbit far from hospitals and banks, while a data center sits on the territory of a state that may not be party to the war and serves customers with no connection to it. Lawyers advising technology firms warn that logical separation of military and civilian data offers little protection if both share the same physical equipment. In the Gulf, it was the host countries that absorbed the damage.
‘Not only an American problem’
The reserve itself is not limited to American firms. EOS is Australian and NorthStar is Canadian. States are responsible for the companies they license, and scholars have already asked whether a company’s role in a conflict could compromise the neutrality of its home state.
The reason governments accept this exposure is largely financial. Frontier AI requires enormous amounts of computing power, energy and chips, and a dedicated military version of that infrastructure is beyond the budgets of most states, including wealthy ones. The savings appear immediately in a defense budget, while the risk appears later, on a company’s balance sheet and over the airspace of a host country, where it has never been priced.
‘Sorting the risk’
One way forward is to stop treating ‘commercial’ as a single category. At present a satellite or a data hall is either ordinary commercial infrastructure or it is not, and the contract rarely says which. Assets could instead be sorted into three tiers according to how much military work they carry. Ordinary commercial capacity would sit at the bottom, with no military workloads and no special obligations. The middle tier would hold capacity that a provider has openly designated for military use, in dedicated halls, regions or spacecraft, backed by government war-risk indemnification and agreed rules on activation and on cutting off other customers. At the top would be government-owned, hardened infrastructure for the few functions that cannot afford any doubt.
The middle tier matters most. If purpose and use are what make an asset a target, it is better for them to be declared openly than hidden inside shared buildings and shared constellations. A designated asset can be hardened, defended and insured, and its host or licensing state can consent in advance. Maritime war-risk insurance already works on a similar principle, in which danger is declared and priced.
There are fair objections. Someone must decide which tier an asset belongs in, providers may game the labels, and a military designation could make an asset a more attractive target. However, the current approach does not avoid these problems, but rather leaves them unaddressed.
Governments want commercial speed and commercial prices, and they are getting both. How this entanglement will work once fighting begins remains unanswered, and hosts, companies and militaries need to settle who pays before that day arrives.
Author: Moneeb Mir, Associate Research Officer, Center for International Strategic Studies, AJK.