In May 2026, four Russian military satellites came close to ICEYE-X36, a private synthetic aperture radar satellite widely believed to be providing battlefield intelligence for Ukraine, in an apparently aggressive and risky "grey zone" maneuver that spooked many in the space industry. Close enough to jam or interfere with the satellite's signal, blind it with a laser, fire a metal dart at it or, simply, force it to maneuver and burn fuel, the satellites' seemingly intimidatory behavior has increased the West's desire to develop defensive technology for satellites. In October 2026, it was reported that one of those same four Russian satellites passed within 500 meters (1,650 feet) of the same ICEYE satellite.Now, U.K. startup Shield Space is racing to apply the lessons of drone warfare in Ukraine to develop swarms of up to 100 very small, low-cost, attritable drone-like autonomous spacecraft called cubesats. Released by a car-sized carrier in Earth orbit, these microwave-sized spacecraft will act as interceptors to protect Western satellites, and their deployment will mark a next step in the evolution of orbital warfare. Shield Space says that these space drones will use "effectors" like lasers or microwaves, or "kinetic standoff" techniques, to defend Western satellites even if controllers lose contact with them."We're talking about deploying 100 cubesats on a protect-and-defend mission for less than £100 million [$134 million US], which is how [much] you would pay for one satellite previously," said Graeme Ritchie, co-founder and CEO of Shield Space.Launched in 2023, NASA's civilian project Starling demonstrated that a swarm of four small satellites can cooperate and operate autonomously in space on a civilian science mission without relying heavily on real-time commands from ground control. But that seems like a far cry from the adversarial or defensive missions Shield Space and its competitors imagine. It is telling that the leading space powers appear not to have demonstrated swarm combat in space — although they tend to be reluctant to reveal their space warfare capabilities.In the fourth quarter of 2027, Shield Space plans to take this next step in space warfare's evolution by demonstrating rendezvous and proximity operations (RPO) with two very small spacecraft. This effort, named Project Broadsword, will launch on SpaceX's Transporter-21 rideshare mission. It will send up an 8U cubesat and a smaller 3U target craft, which Shield Space plans to maneuver from 100 kilometers (62 miles) apart to within 100 meters (330 feet) of each other and, through four stages of autonomy, from human-in-the-loop to "approximately" fully autonomous. (A cubesat unit, or U, is a cube that measures 10 cm, or about 4 inches, on a side.)Russia's Luch-2 satellite approaches another geostationary satellite on April 12, 2024, as tracked by the French company Aldoria's observation stations. (Image credit: Aldoria)The gaps between a true satellite-defense capability and this demonstration appear to be great, in terms of the technological and practical challenges of squeezing propulsion and power systems, AI compute power, sensors, the lasers the swarm needs to communicate and weapon systems onto small, lightweight maneuvering satellites. But it is a gap that Shield Space believes it can bridge through the purchase of relatively cheap off-the-shelf technology rather than more expensive bespoke designs."Yes, it's a problem because of size, weight and power," Ritchie told Space.com. "We've had to totally innovate from scratch some elements of the mission… but we have only got roughly 100 cm by 100 cm [40 by 40 inches], so saving space or having systems that are already completely integrated is the main aim.""We have commissioned EnduroSat to build us an autonomous 8U cubesat that has chemical propulsion and its own imagery-collecting capability, which we will put our software on," said Dan Molland, co-founder and CTO of Shield Space, "and which we will then use to conduct our RPO. "The good thing about cubesats is they're very standardized and they're very off the shelf," he added. "There's always some level of adjusting you've got to do, but these cubesats are a lot more plug and play than more conventional platforms, which are usually all custom-built."Illustration of two Shield Space cubesats in orbit. (Image credit: Shield Space)Yet to focus on the technical issues may be to miss the point. "It takes an extraordinary amount of data to train the models and algorithms to achieve a fully autonomous mission," said Ritchie, "and the problem [unlike in the Ukraine war] is that data is not readily available. "We can train AI using synthetic data, but it's just not quite good enough for space. It's incredibly important to gather data from missions like this if we really want autonomous missions in the future." There is also the complexity of the AI needed in a war-fighting system. "Starling is a really great example," Molland told Space.com. "With Starling, you're essentially just asking a guidance system to make sure you stay within a given range of this or that satellite… But if you are controlling a swarm to deliver [a] military mission, you have a much more complex set of questions."It also comes with the challenge of ensuring that AI complies with any rules of engagement.But Ritchie and Molland believe these technical challenges can be overcome, and they have an aggressive schedule of demonstrations planned that they think will prove it.Bleddyn Bowen, associate professor of astropolitics at Durham University in the U.K. and author of the book "War in Space: Strategy, Spacepower, Geopolitics" (Edinburgh University Press, 2020), is more cautious about Shield Space's plans for the evolution of space warfare. "They remind me of a late Cold War missile-defense scheme called Brilliant Pebbles," he told Space.com. "It was one of the more advanced research and development programs in the Strategic Defense Initiative."The plan was to defend the U.S. against Soviet intercontinental ballistic missiles (ICBMs) by using thousands of small autonomous orbiting interceptors to destroy them by collision."With enough investment, a lot is feasible," Bowen said. "But you're looking at many years of development and testing, solving the problem of testing kinetic weapons in space without leaving debris, and finding someone to buy the system." It is clear from Russia's ongoing invasion of Ukraine that the future of warfare includes swarms of low-cost autonomous drones — and Shield Space is already planning an aggressive schedule of missions to prove that this pattern applies to the final frontier as well. According to Ritchie, the startup hopes to quickly go from the first RPO demonstration to testing two full-sized 16U cubesats. It will then move on to a full carrier of 10 such cubesats, and then to a "super carrier" of 10 carriers, which will be deployed across multiple orbits. "We then get towards our future vision, which we call space-based carrier strike," said Ritchie. In this scenario, "responsive launch or pre-positioned carriers will give us an ability to protect and defend entire orbits, should you choose."