Last updated: 6/9/2026
The year is 2030, you’re doing recon in the mountains of [REDACTED] in pursuit of a dangerous group. Trekking through the middle of the night in complete darkness, you wear a futuristic looking helmet using the worlds most advanced night vision to let you see as day. A notification sounds. An alert that a satellite has detected UAVs swarming towards your position. They’re instantly placed within your heads-up display, highlighted in crimson over the dusky horizon. Your hand instinctively moves up to press the zoom button but a friendly interceptor has already fired an airburst missile and eliminated the target before you get to see any action. A complex system of edge nodes has sensed and acted from satellite to decision point to launcher to interceptor so fast that an un-enhanced individual would have never noticed the threat.
You are an American. Fear is only part of the Chinese lexicon.
I saw a cool announcement today from Anduril about observing a SpaceX Falcon Heavy second stage in orbit, revealing a massive network of SSA (space situational awareness) telescopes and wanted to know more about their space program so I made a quick writeup.
Back now in 2026. Private defense companies are scrambling to construct networks of complete battlespace awareness spanning all domains: sea, air, land, subterranean!? and space. Some of these companies speak louder than others. Few of these modalities are mentioned as quietly as the space domain. Hiding beneath the endless triage of slop-generating orbital datacenter concepts Anduril has not been working on generating bikini pictures from space. Instead, they’ve been working to establish domain and decision dominance from 'space to mud' as Palantir puts it. This article will briefly cover their acquisitions, current public programs, and how it all fits together.
Anduril was not founded as a space company but inclusion of the celestial domain is unavoidable in the pursuit of complete awareness. Acquisitions indicate which skillsets and domain areas are prioritized for the mission
American Infrared Solutions (AIRS): the hardware stack of a sensing layer. AIRS provides high-end cooled long range and long wave thermal imaging cameras. Thermal imaging cameras are historically finicky and low resolution. Improving upon them requires extreme sensitivity and active cooling. The colder an object is, the longer of an infrared wave you have to sense to track it. Near infrared (NIR) and short-wave infrared (SWIR) sensors are becoming more common since the bands are close to visible light and similar to standard hardware can be used. Unfortunately, these bands are mostly good for detecting items operating at hundreds of degrees. Once your enemy begins cloaking their exhaust plumes or operating smaller vehicles you need to detect longer waves.
Exoanalytic Solutions: the software stack of a sensing layer. Exoanalytics focused on tracking and characterizing space-borne threats. Government contracts included modelling and tracking re-entry of orbital objects. Development of space situational awareness algorithms to anticipate the behavior and assess the capabilities of other resident space objects (RSOs.) Tools like InterceptAI optimize the interception of hypersonic glide vehicles. FATES allows users to easily select between multiple interception or weapon options. Various custom LLMs make all of this information accessible to the end user.
What they haven’t acquired: edge compute, the processing stack of a sensing layer. Following a brief build vs buy study of the hoard of space-based compute providers, Anduril most likely discovered that everyone in low Earth orbit (LEO) is already using COTs Jetson GPUs and their current setup will do just fine.
What does this make? A 2–400-person team developing an end-to-end solution that detects, identifies, and tracks threats from and in space. This information can then be routed back through Anduril’s Lattice AI platform to give users a superior level of awareness. The Golden Dome playbook. Perhaps they will be a big contract winner, if the DoW ever figures out what is going on with that program…
Multiple space-based demonstrations are planned for 2026 spanning from low to geostationary orbits.
Impulse Space Partnership: a thermal imager riding on a Mira spacecraft will navigate from LEO to GEO in under a day to autonomously approach, inspect and characterize RSOs. This simulates (or emulates…) the capability to quickly identify the capabilities and behaviors of foreign national assets in space. What are their payloads, what are their communications, their vulnerabilities and where are they pointing? This dance of satellites maneuvering to inspect each other will become increasingly common in the near future.
The bus itself is a espa-class micro-satellite with a payload capacity of hundreds of kilograms and up to 850 m/s delta-V maneuvering capability. Its marketed as ‘the agile spacecraft’ of Impulse’s line. Communications are limited to 4Mbps down and 400kbps up so it is likely that Anduril will be implementing edge processing onboard to prioritize the downlink of condensed and highly characteristic data.
The Apex Aries Bus: an opportunity to demonstrate the capabilities of Anduril’s sensors. They integrated thermal imagers and edge processing hardware onto an Apex Aries bus to demonstrate generating low-latency intelligence from observing other objects in space. Infrared (IR) is an odd choice for broad sensing of objects in orbit due to the aforementioned poor resolution. RSOs would almost certainly appear as sub-pixel objects within an IR camera. It is possible to gather a lot of information such as position, orbit, material, size and spin rate from single pixel glints (see SpaceX StarGaze) but results will improve even within single-pixel systems as the pixel size falls. For this reason, it is likely that Anduril is also taking this opportunity to experiment with more traditional monochrome sensing imagers. This will almost certainly be utilizing the high-power configuration of the Aries bus since the base 175 watt maximum generation, ~50-75 watt orbit average power would be on the lower end of sufficient for in space processing. There are cubesats these days beating those numbers. When we account for the fact that the processing hardware is likely to be based on terrestrial, less power constrained, designs we can also assume that the bus will have tracking solar panels to get closer to the 500W maximum generation in a LEO orbit where the sun may not always present at a fixed angle. Communications on this bus are far less limited than those in the higher up geostationary experiments - reaching up to 125 or 350 Mbps dependent on configuration. Propulsion is likely to be required to meet de-orbit regulations. With launch and operations, bus costs may exceed ten million dollars. Quite a significant investment and a clear indicator that Anduril plans to pursue the space ecosystem aggressively.
The Argo Hail Mary: This is another inspection mission where Anduril is placing their sensor suite on a highly maneuverable satellite but to dip in and out of LEO and towards higher-value geostationary targets. The overall mission design appears to be very similar to the Impulse Space demo. This will be Argo’s first hardware in space and they are packing it with a variety of payloads from Anduril, ThinkOrbital, and Infinite Orbits. The fact that the bus will be using KULR batteries and has unrealistic early-stage renderings (where they still opted to blur 'technical' details?) on their website makes me think this company is a LOT less technical than the layman would realize. I’d imagine Anduril is getting this payload mass up VERY cheap or for free in exchange for legitimizing the company.
An aside on KULR for those out of the know – KULR is a company that tried reselling standard battery cells to NASA by lot testing them and marking up the price. When that didn’t work too well - they pivoted to being a bitcoin holding company. Somehow the markets loved that move and allowed KULR to dilute the ever-living-crap out of shareholders increasing the number of shares 4x to raise $123M on what is now a $170M market cap. A play somehow more impressive than the utter shafting that Sidus gives their shareholders every time the stock rises more than 10% in a week… This is the absolute bottom tier of new-space sludge. It is possible, but doubtful, that they used some of their crypto-scam money to hire competent engineers.
Anduril announced (today!) that they have a network of over 400 ground telescopes around the globe executing similar space situational awareness missions and collecting data to be integrated with Lattice. This is more observation points than players like LeoLabs or Slingshot Aerospace have combined. A very impressive quiet buildup making them one of the most knowledgeable groups pertaining to events occurring in orbit. Telescopes instead of radar is an interesting choice. For SSA radar is typically a better fit for mass observation and cataloging of items due to wider fields of view and better tradeoffs on resolution through the atmosphere. Telescopes are typically used for more targeted efforts meaning Anduril is focused primarily on tracking known objects of importance. This matches the ethos of their orbital efforts wherein Anduril is likely focused on tracking and characterizing the behavior of foreign national assets in space.
That said… if they ever want to go commercial and offer a solution to the SpaceX Transporter TLE lottery problem many paying customers would surely welcome them.
The overall theme of Anduril’s work in space is clear. They are not interested in Earth Observation (EO) which is better played out in other hands. They will not primarily sell payloads or components to bus manufacturers. There is no data center play. Anduril is here to carefully track potentially malicious objects in space – aggressively to the point of chasing them down - then route that information into Lattice so that other systems can act on the threat. This is the sensing, tracking or ‘custody’ layer of the Golden Dome system. If a foreign satellite is watching, scanning, listening or otherwise targeting American service members Lattice will know. If a missile exits the atmosphere, Lattice will know. This information can then be forwarded onto other systems like Palantir’s Foundry or Gotham (as announced in 2024) for fusion with other sources before triggering a countering system. The satellite detects the threat, Gotham issues confirmation, and defensive actions are taken. The initial scene plays out.
Anduril recently contracted with the Space Force to develop space-based interceptors (SBIs) as a component of Golden Dome. These interceptors will primarily target hypersonic or ballistic missiles operating beneath the orbital regime. Pre-positioning distributed interceptors in space allows for reduced interception time. See the age old classic 'Brilliant Pebbles' initiative [https://en.wikipedia.org/wiki/Brilliant_Pebbles]. Interesting to note here is that the architecture of such systems – low orbit, proliferated, disposable – is entirely opposing their current public mission set riding on expensive geostationary buses. There is no off the shelf solution here. Anduril will have to design these vehicles in-house. Eventually they will have to learn to make satellites.
For now, partnering with K2-Space and using their 'Mega class' satellites as orbital carriers once again allows Anduril to skip many of the challenges associated with long-term satellite design and operations.
Worth noting is that Anduril may not be skipping the bus designs to avoid difficult problems but instead using existing solutions to skip the typical years-long experimental phase that space companies go through before fielding operational systems. We need Golden Dome yesterday and adding a few more years to the timeline would be unacceptable. The ideal system is still a network of highly distributed nodes but a perfect system never delivered is worse than a limited system that actually gets made.
Similar to their ambitions of mass proliferating a system of autonomous weapons in the sea, the same can be done for interceptors in space. This is by far the most interesting work in progress, but also the one we will hear the least about.
A final look at Anduril’s career page shows openings for space tracking software engineers, space imaging specialists, guidance navigation and control engineers, space computer vision programmers, flight software developers, space battle management, and even a chief engineer of space (👀) primarily in Costa Mesa California, Boulder CO, Lexington MA, and of course Washington DC. These roles all follow in line with what we would expect from a company developing a complex space situational awareness network.
Last note: yes, Anduril also produces solid rocket motors. These can absolutely be used extra-terrestrially but I know little on the topic and will stick to satellites.