The short version

• The cross waiver is the defining feature of a launch services agreement. Everyone bears their own losses regardless of fault, and the waiver flows down the entire contract chain.
• US law requires it. 51 U.S.C. 50914 and 14 CFR 440.17 make reciprocal waivers a condition of a licence.
• Above the waiver sit three tiers: mandatory insurance sized to maximum probable loss, government indemnification up to $1.5 billion indexed from 1989, and then the operator again, uncapped.
• The indemnification tier has never been triggered and Congress has never appropriated against it. Its current expiry is 30 September 2028.

A launch failure is the rare commercial event that can generate claims in every direction at once. The launch provider’s vehicle destroyed the customer’s satellite. The satellite’s failure damaged the pad. A component supplier three tiers down may have caused all of it. Government range equipment was involved. Several insurers are subrogated to several of those claims.

Litigated conventionally, one failure produces a decade of circular claims in which every participant sues every other participant and most of the money goes to lawyers.

The industry solved this in the 1980s and has not changed the solution since.

What a cross waiver does

A cross waiver is a reciprocal, no-fault waiver of claims. Each participant in a launch agrees to bear its own property damage and its own employees’ injuries, and waives claims against every other participant, regardless of fault.

The participants are drawn widely: the launch provider, the customer, their contractors and subcontractors at every tier, crew and spaceflight participants, and the US government. Each party must flow the waiver down its own contract chain, so a fourth tier supplier is inside the ring even though it has no contract with the launch provider.

The effect is to convert an interlocking commercial and governmental supply chain into a set of self-insured silos. Each party insures its own property, prices that into its contract, and stops worrying about who was at fault.

Two standard exceptions survive. Wilful misconduct, and claims exceeding the required insurance amounts.

The cross waiver does not touch third party claims by uninvolved people on the ground. Those are handled by the insurance and indemnification tiers below.

Where the requirement comes from

This is not purely a matter of negotiation. US law mandates it.

51 U.S.C. 50914(b) requires a launch or reentry licence to contain reciprocal waivers among the licensee, its contractors, subcontractors, customers and spaceflight participants, each assuming responsibility for its own employees’ injuries and property losses. It also requires the Secretary of Transportation to negotiate reciprocal waivers protecting the government, its agencies and its contractors, except for claims exceeding the required insurance.

14 CFR 440.17 supplies the detail. Waivers are required between the licensee and its contractors and subcontractors and the first tier customer and its contractors and subcontractors, with spaceflight participants, and with crew.

An operator that wanted to negotiate this away could not. It is a licence condition.

The three tiers above the waiver

Third parties, meaning people who are not participants in the launch, are not covered by the cross waiver. For them there is a structure.

Tier one: mandatory insurance sized to maximum probable loss

Under 51 U.S.C. 50914(a) a licensee must obtain liability insurance or demonstrate financial responsibility to cover maximum probable loss from third party claims and from claims for damage to US government property.

MPL is defined at 14 CFR 440.7 as the greatest dollar amount of loss for bodily injury or property damage reasonably expected from the licensed activity. The FAA issues a determination within 90 days of a complete request.

The probability thresholds are the technical heart of the regime. One in ten million for third parties. One in one hundred thousand for government personnel and property. Losses less probable than those thresholds are excluded from the calculation.

Statutory caps sit above the calculation: $500 million for third party liability insurance and $100 million for government property, or the maximum available on the world market at reasonable cost if that is lower.

Duration matters too. Launch cover runs until 30 days after payload separation, and reentry cover for 30 days from initiation of reentry. Certificates of insurance are due at least 30 days before activities begin.

Tier two: government indemnification

Under 51 U.S.C. 50915 the government may pay successful third party claims above the required insurance, up to $1.5 billion plus amounts reflecting inflation after 1 January 1989.

The qualifier is decisive. Payment is available only to the extent provided in advance in an appropriation law or to the extent additional legislative authority is enacted. Congress has never appropriated against it and it has never been triggered.

Tier two is therefore a promise to consider paying, not a guarantee of payment. Anyone modelling it as a hard backstop is modelling something that does not exist.

Tier three: the operator, uncapped

Claims above the indemnification ceiling fall back on the licensee, with no cap. In practice a loss of that size would be an insolvency event.

The expiry date to diarise

Under 51 U.S.C. 50915(f), the indemnification regime applies to a licence issued or transferred for which the Secretary receives a complete and valid application not later than 30 September 2028. That date was set by Pub. L. 118-159 on 23 December 2024, extending it from 30 September 2025.

Note the trigger precisely, because secondary commentary garbles it regularly. It is the date the FAA receives a complete and valid application, not the date of the launch. A licence applied for before the deadline carries the regime with it.

Given the history of short serial extensions, plan on it being extended again and do not build a business case on the assumption.

Drafting points that matter

Flow down is the weak link

The cross waiver only works if every participant is inside it. The commonest failure is a subcontract deep in the chain that omits the flow down, usually because it was papered on a supplier’s standard terms for a component that nobody thought of as launch related. Audit the chain, do not assume it.

Define the participants by reference, not by list

Personnel and suppliers change between signature and launch. A defined class that captures contractors and subcontractors at every tier is more robust than a schedule of named entities.

Wilful misconduct is narrower than it sounds

The exception is not gross negligence and it is not recklessness in the loose sense. Parties who assume it gives them a route back into the ring after a failure are usually disappointed.

Insurance and the waiver must line up

If the cross waiver allocates a loss to a party whose policy does not cover that loss, the waiver has created an uninsured exposure. This is a coordination problem between the contract team and the insurance broker and it is where most real risk sits. We look at the market side in our piece on space insurance.

Non-US launches follow a similar shape

European launch contracts through Arianespace, and licences under the UK Space Industry Act 2018 or India’s IN-SPACe regime, use comparable no-fault allocation and insurance requirements, though the numbers and the indemnity limits differ. The UK, for example, works to a standard €60 million insurance requirement and indemnity limit for typical orbital operations, and the Space Industry (Indemnities) Act 2025, in force from 18 February 2026, now makes a liability cap mandatory in every operator licence rather than discretionary.

The short version

• The launch and in-orbit insurance market is small. Premium income was over $650 million in 2025 against claims of around $503 million.
• 2023 broke it. $1.43 billion of claims against roughly $550 million of premium, which turned the rolling five year margin negative for the first time since 2001.
• Capacity of about $160 million left the market in 2024 and rates have stayed elevated since.
• Twenty five insurers were writing space risk in 2026, with the UK accounting for roughly 43 percent of launch capacity.

Space insurance is one of the smallest specialty lines in the world and one of the most consequential. It sets the floor under launch contracts, it decides whether a satellite programme is financeable, and it is priced by a few dozen underwriters who all know each other.

It is also, on the numbers, a business that has struggled to make money.

The size of the pool

Aon’s Space Insurance Market Report is the standard reference, and its Q1 2026 edition gives the shape of the market. All figures are in US dollars and claims are measured by year of occurrence, so they get revised as settlements complete.

Two things stand out. The whole global launch and in-orbit premium pool is smaller than the annual revenue of a mid-sized regional insurer. And a single bad year can wipe out several good ones.

Note also the definitional trap. Market research reports circulating in 2026 quote a “space insurance market” of around $4.43 billion growing to $6.23 billion by 2030. Those figures use a much broader definition of the sector and are not comparable to the Aon numbers. For launch and in-orbit premium specifically, the $550 to $650 million range is the one to cite. Do not blend the two.

What 2023 did

2023 is the pivot year that explains current pricing. Claims of $1.43 billion against premium of approximately $550 million produced a gross margin of minus $882 million. That single year took the rolling five year gross margin negative for the first time since 2001, to minus $643 million.

The response was capacity withdrawal. Roughly $160 million of launch and in-orbit capacity left the market during 2024.

Recovery since has been real but thin. Profitable 2024 and 2025 restored the rolling five year margin to plus $62 million at the start of 2026, which is barely above break-even across the cycle. That is the context for every renewal conversation an operator is having.

2025 in detail

Premium growth in 2025 came from several high value insured geostationary launches, launch vehicle flight-only covers and steady in-orbit renewals.

The claims side, at roughly $503 million against premium above $650 million, gave a loss ratio of 77.4 percent. That is a profitable year in a line with high volatility, but not a comfortable one.

The individual losses tell you what the market actually pays for: two small Chinese launch failures, approximately $30 million each for Astranis UtilitySat, which suffered a propulsion failure in transfer orbit after a December 2024 launch, and MethaneSAT, which suffered a power failure, and the potential total loss of SPAINSAT NG-II after it was struck by a space particle en route to its final orbital position following an October 2025 launch.

That last one is worth pausing on. A single high value geostationary spacecraft lost to an in-orbit event absorbed the year’s premium growth. In a pool this size, the loss distribution is dominated by a handful of large risks.

Who writes it

Twenty five insurers were underwriting space risks in 2026, the same number as 2025 but with real churn beneath the headline.

Phemis Underwriters, formed from the former Hiscox space team, began writing in mid-2025 with $25 million of capacity. Aesir Space, from the former Volante team, resumed in 2025 with $8.5 million. Hive Underwriting was new for 2026 with $25.5 million. Occam paused in November 2025 and resumed in February 2026 with replacement capacity from Nexus Underwriting and a reduced $18 million line.

Theoretical capacity for 2026 was $709.4 million for launch and $669.4 million for in-orbit, excluding Chinese markets. The geographic concentration is striking: the UK accounts for roughly 42.7 percent of launch capacity at $303 million, with Asia at $160.7 million, then France, the United States and the rest of Europe.

What actually gets covered

Pre-launch

Damage during manufacture, testing and transport to the launch site. Usually a property policy rather than a space policy, though the transition point to launch cover needs care.

Launch

Typically attaching at intentional ignition and running through to a defined point after separation, often including initial in-orbit testing. This is where the concentration of risk sits and where rates are hardest.

In-orbit

Annual cover for the operational life, renewable. Responds to total loss, constructive total loss and partial loss, with partial loss usually measured against a defined capacity or transponder metric.

Third party liability

Cover for damage to people or property outside the launch participants. For a US licensee this is sized to the FAA’s maximum probable loss determination, capped at $500 million by statute. We set out how that interacts with the contract in our piece on cross waivers and indemnification.

Why the cross waiver matters to the underwriter

The cross waiver in the launch services agreement is not just a contractual nicety. It determines what the insurer is actually on risk for.

Because each participant bears its own property loss regardless of fault, an in-orbit policy on the satellite responds to the satellite’s loss and has no subrogation route against the launch provider. The premium reflects that. If a contract were negotiated without an effective flow down, the risk profile changes and the underwriter needs to know.

This is the most common practical failure we see described. The contract team and the broker work from different assumptions, and the mismatch only becomes visible after a loss.

What operators should expect at renewal

Aon’s own guidance for 2026 is that hard market conditions persist through the early part of the year, with possible easing later if profitability holds. Placements with large sums insured remain difficult and underwriters are described as extremely selective.

Practically, three things follow. Large single risks will need to be spread across more markets than they were five years ago, which means more brokers, more disclosure and more time. Loss history and heritage matter more than they did in a soft market, so a new bus design without flight heritage will pay for it. And the flight-only structures that grew in 2025 exist because they let underwriters take a defined slice rather than the whole exposure, so they are worth discussing early rather than treating as a fallback.

The short version

• Orbital slots and spectrum are not owned. They are recorded, and the record is held by the ITU Radiocommunication Bureau.
• A filing has to be brought into use within seven years, and for both GSO and NGSO that means a satellite maintained on the notified assignment for 90 continuous days.
• Since WRC-19, NGSO constellations also face milestones of 10, 50 and 100 percent deployment at two, five and seven years after that.
• Missing a milestone does not cancel the filing. It shrinks it, which is the ITU’s answer to paper constellations.

Every satellite communications business rests on two assets it does not own. A position in the sky and a slice of spectrum. Neither is property. Both are entries in an international register maintained by the ITU Radiocommunication Bureau in Geneva, and the entries are made by national administrations, not by companies.

That structure explains a lot of behaviour that otherwise looks strange, including why an operator with no obvious connection to a small state files through it.

The filing chain

The process runs through the Radio Regulations, principally Article 9 for coordination and Article 11 for notification.

Advance Publication Information

Historically the first step, in which an administration told the Bureau what it intended to operate and the Bureau published it. Since WRC-19 the API stage was largely folded into the coordination request for most fixed satellite, broadcasting satellite and mobile satellite filings, so in practice most commercial systems now start at coordination.

Coordination request

Filed under RR No. 9.30. The Bureau publishes the request, identifies which administrations may be affected, and the notifying administration then has to coordinate bilaterally with each of them. This is where most of the time and most of the negotiation goes. There is no arbitrator. If two administrations cannot agree, the filing sits.

Notification and recording

The assignment is notified to the Bureau, examined for conformity with the Table of Frequency Allocations and the applicable coordination, and if the finding is favourable, recorded in the Master International Frequency Register. Recording is what confers the right to international recognition and protection from harmful interference.

Bringing into use

The assignment must be brought into use within seven years of receipt of the complete coordination request. Miss it and the filing is cancelled.

For a geostationary network, bringing into use means a space station with the capability to transmit or receive on the notified frequency, maintained at the notified orbital position, for a continuous 90 day period. For a non-geostationary system, it means deploying a satellite in one of the notified orbital planes and maintaining it for 90 continuous days.

The problem with 90 days and one satellite

Read the NGSO rule again. One satellite, 90 days, and a filing for a constellation of ten thousand is brought into use.

That was the position before 2019, and it produced exactly the behaviour you would expect. Filings for enormous constellations were made and held with minimal deployment, blocking coordination for anyone who came later. The industry term is paper satellites.

WRC-19 fixed it with Resolution 35, which added a milestone based deployment schedule measured from the end of the seven year bringing into use period.

Administrations must report deployment to the Bureau no later than 90 days after each milestone period expires.

The sanction is elegant. It is not cancellation. It is automatic reduction of the recorded satellite count. Fall short of the 10 percent milestone and the filing is capped at ten times the number actually deployed. Fall short of the 50 percent milestone and it is capped at twice the number deployed. Fall short at 100 percent and it is capped at exactly what is flying.

An operator that files for 10,000 and deploys 400 by milestone one keeps a filing for 4,000, not 10,000. The paper constellation shrinks to the size of the real one.

Why the flag state question arises

Filings are made by administrations. A company does not file. It asks a national administration to file on its behalf, and that administration then bears the coordination workload and any international consequences.

Administrations differ in how quickly they file, what they charge, how they prioritise between competing domestic applicants, and how vigorously they coordinate. Those differences are real and they are a legitimate factor in structuring, in the same way that vessel registration is in shipping.

They also feed the criticism that the system rewards administrative arbitrage rather than technical merit. That criticism has not produced a change in the rules, because the rules are made by the same administrations.

WRC-23 and what is coming at WRC-27

WRC-23 ran in Dubai from 20 November to 15 December 2023, with roughly 3,900 delegates from 163 Member States. Three outcomes matter for NGSO operators.

New orbital tolerances were introduced. For systems with eccentricity below 0.5 and apogee under 15,000 km, apogee, perigee and inclination tolerances are 70 km during deployment, tightening to 30 km afterwards.

A regulatory framework for Ka-band satellite to satellite links was approved, which lets LEO constellations relay through GSO and MEO systems.

On sustainability, the Radiocommunication Assembly adopted Resolution ITU-R 74 on sustainable use of spectrum and orbital resources, and the ITU began requesting post-mission disposal information from NGSO operators.

The flashpoint was equivalent power flux density. NGSO operators wanted a WRC-27 agenda item to revisit the EPFD limits that protect GSO networks, which date from a period when the NGSO population was tiny. GSO operators and several administrations opposed. EPFD revision was kept off the WRC-27 agenda, with technical studies continuing for a report without regulatory action.

For WRC-27, the items to watch are:

• AI 1.5, measures to limit unauthorised operation of NGSO earth stations and to allow a country to be excluded from an NGSO system’s service area. Both are technically contested.
• AI 1.16, NGSO interference into radio astronomy, including possible recognition of radio quiet zones for ALMA in Chile and the SKA in South Africa. This is now the main dark and quiet skies battleground inside the ITU.
• AI 1.12 to 1.14, new mobile satellite allocations including direct to device.

The gap the ITU does not fill

It is worth being clear about what this system does and does not do. The ITU allocates and records spectrum and orbital positions. It does not license satellites, it does not regulate debris, it does not adjudicate collisions and it has no enforcement arm.

An operator with a perfect ITU filing and no national authorisation cannot launch. An operator with both can still create a hazard nobody has authority to stop. That gap is the subject of our piece on space traffic management.