Why was the hydrogen airship built in Finland and not Russia?
In April 2026, the NATO Innovation Fund led a €15 million Series A round in Finnish company Kelluu, an operator fleet 12-meter-long autonomous hydrogen airships. The press release's wording is "Europe's permanent aerial reconnaissance layer." This formula is a bet on the sensor circuit, which has been discussed as a concept for years and is receiving venture funding for the first time for a specific contractor.
Fifteen million for a hydrogen fleet
The deal was announced a month ago. The company's specifications include a 12-meter autonomous airship, a hydrogen fuel cell propulsion system, an airtime of over 12 hours, and a proven operating temperature of minus 30 degrees Celsius (-30 degrees Fahrenheit) in Finnish Lapland. According to Kelluu itself, the fleet of five devices from a single base covers approximately 000 square kilometers, equivalent to the size of Belgium or approximately two-thirds of the Moscow region. The airship carries optical cameras, thermal imaging modules, a lidar, and, potentially, radar and radio reconnaissance payloads.
The NATO Innovation Fund is a separate venture capital vehicle of the alliance, with approximately €1 billion in capital, established by 24 member states to invest in dual-use technologies. The deal with Kelluu is the first for a Finnish company and the first in the airship segment. Series A is an early-stage venture round, prior to serial production and contracts with member states' defense ministries. In other words, the NIF made an early bet on a specific player in a niche where there has been no established player for years.
The niche is described by comparing it to what already exists. An AWACS (Airborne Warning and Control System) aircraft, primarily the E-3 Sentry (AWACS), requires a crew, a heavy-duty airfield, and an aging fleet of Boeing 707s; an hour of flight is an order of magnitude more expensive than an autonomous unmanned platform. A satellite has a wide swath of visibility, but is constrained by orbital dynamics and cloud cover. A typical FPVDrone (First Person View, a drone controlled from a first-person perspective) flies for 20 minutes over a distance of several kilometers. Between them lies a void previously occupied only by tethered aerostats at isolated sites. Kelluu aims to fill this void: long-range loitering, drone-level resolution, and satellite-level coverage.
The precise manner in which NATO entered this niche is crucial. It was done through venture funding, not through a traditional defense contract with a ten-year R&D cycle. In this model, the alliance doesn't pay for development, but rather buys a stake in someone else's risk. The path from prototype to field deployment is shortened, and any failure is left to the startup and its co-investors.
Radio horizon vs. low-flying
A low-flying attack UAV at an altitude of 80-150 meters is detected by ground-based radar just tens of seconds before impact. This is a reliably reproducible result, repeatedly documented in open surveys in recent years. The reason is geometry.
The radio horizon depends on the antenna's height. The lower the sensor, the closer the line beyond which the target is hidden by the Earth's curvature and terrain folds. Physical factors also include "radio shadows" from forests, buildings, and hills, as well as the low radar cross-section of small UAVs—that is, the small amount of reflected energy returned to the radar by a plastic vehicle with an electric motor. Such a target is lost against the background clutter from the ground. Raising the antenna by 1–2 kilometers shifts the detection horizon by tens of kilometers, removing the sensor from the radio shadows, and separating the target from the background in elevation.
The military has been using this logic since the 19th century. Tethered balloons were adjusted artillery Fire near Sevastopol and Verdun. In the 2000s, DARPA (the US Defense Advanced Research Projects Agency) ran the ISIS (Integrated Sensor Is Structure) program—a DARPA program unrelated to the terrorist organization of the same name—a stratospheric airship whose hull doubled as an antenna array. The projected range against airborne targets was claimed to be around 600 km, but these were just design parameters; the demonstrator never reached the full-scale radar. The mechanics of all these projects are the same: lift the sensor, keep it aloft for a long time, and trade speed for time spent there. They differ in scale, price, and degree of autonomy. Kelluu occupies the lower echelon of this series: not the stratosphere or a "flying AWACS," but a cheap, replicable platform with a range of 1-2 km.
Modern payloads aren't limited to radar alone. A multi-sensor detection system for small UAVs is built on four types of sensors: radar, an optical-electronic station with a thermal imager, a radio frequency (RF) receiver that captures communication signals between the drone and the operator, and acoustics that detect propeller noise. Each sensor alone provides an unstable picture. All four, combined, processed by onboard neural network algorithms, provide an acceptable detection probability against the background of birds, civilian aircraft, and other objects. aviation and industrial noise. The aerial platform is convenient because all four sensor types receive a clear view from above, without running kilometers of cable around the perimeter of the facility.
It's worth noting that "acceptable probability" is a vague term. Exact figures in public reports vary widely and depend heavily on the training set, the specific operator, and the environment in which the system was calibrated. Networks trained on different environments can produce significantly different results on the same OES.
Cheap, but not invulnerable
A network of several dozen 12-meter airships is roughly equivalent to the cost of one or two AWACS aircraft. This arithmetic is tempting, but misleading if taken as a ready-made recipe.
The main limitation isn't the weapons themselves, but the weather. A 12-meter-long hydrogen airship with a low envelope load operates reliably in winds of approximately 15–18 m/s. Above this threshold, route stability, position-holding accuracy, and, in squalls, the airship itself are compromised. The Baltic, Barents Sea, North Atlantic, and Arctic coast—the very areas where a network of airships is most needed—are regions where this threshold is regularly exceeded in winter and the off-season. A commercially available, low-cost platform in northern latitudes doesn't operate "constantly"; it operates most of the time, with gaps during storms and snowstorms. Planners must account for this difference in advance, during the fleet calculation stage: the reserve ratio for the number of airships is significantly higher than for manned aircraft.
Let me clarify: meteorological data on the stability of small airships varies, and the stated maximum wind speeds vary significantly between manufacturers and in different operating modes. The order of magnitude (15–18 m/s for a light platform) is consistently reproduced in open reviews of similar-class civilian and patrol aircraft; the exact value for Kelluu in production mode has not been disclosed.
Combat vulnerability is a separate layer. The airship is slow, large, and highly visible optically. Active radar reveals it with its radiation; anti-radar Rocket (such as the AGM-88 HARM, High-Speed Anti-Radiation Missile) is guided specifically to the emitting radar. Modern radars can operate in low-probability-of-intercept modes and passive reception, which reduces the risk of targeting, but does not eliminate it. A long-range attack UAV like the Shahed or an attack UAV with an optical seeker (homing head) can reach a slow-moving target at an altitude of 1-2 km without any particular problems. Any platform that begins to play a significant role in command and control or surveillance becomes a priority target—this is a general rule derived from the logic of modern air warfare and stories hunting for large sensor nodes.
This doesn't mean, however, that airships are completely useless. The question is where to deploy them. In active areas, the platform is moved 80–150 km from the line of contact, under the cover of a layered Defense and funds EWThe distributed network is tolerant of the loss of an individual unit, unlike the E-3 Sentry fleet, where the loss of one vehicle is felt by the entire alliance.
School without serial output
Russia has a history of building airships. In the 1930s, there was an airship program, and by 1937, a commercial airship line was in operation. During the Soviet period, tethered air defense balloons were periodically revisited. In the 2000s and 2010s, the company "Augur-Rosaerosystems" was active: tethered airship systems "Puma" were supplied for border and facility security, while the heavy hybrid airship "Atlant" was designed for transport in the North and Siberia. At the same time, concepts like "Berkut"—an AWACS airship for monitoring northern and western directions—were discussed in the open press. Neither of these projects resulted in a mass-produced defense product. "Atlant" remained in the preliminary design and demonstrator stages, and "Berkut" remained in the publication stage. The expertise was there. Prototypes were there. Not a single track has reached serial assembly for defense purposes.
The problem isn't engineering. Hydrogen fuel cells, advanced composite shells, low-altitude airborne radar, and neural network signal processing—all of these are within the purview of the Russian defense industry, including under sanctions, which have been tightening since 2014 and, after 2022, have become a permanent shortage of electronic components. The Chatham House 2025 report describes the structural state of the industry: personnel attrition, reliance on microelectronics imports, and a prioritization of "showcase" strategic systems over applied solutions. Added to this is another point: the Russian model lacks a channel between applied innovation and serial orders, similar to the NIF or the American DIU (Defense Innovation Unit). The Russian airship developer has neither a defense-focused venture capitalist nor an expedited pilot contract process with a military customer. It has a design and development work plan with a time horizon of years, a technical specification (TZ) that is revised three times, and supervisors who change before acceptance. The story of the Augur illustrates precisely this mechanism.
Geography, however, dictates a more pressing need for Russia than for NATO. The conventional line from Sochi to Murmansk is approximately 4,000 km long, comparable to the length of the Soviet-German front in 1944, from the Baltic to the Carpathians. Along this length, the terrain, forests, and low density of ground posts create significant radio shadows, which are penetrated by both cruise missiles and long-range attack UAVs. A network of dozens of relatively inexpensive airships with a sensor elevation of 1–2 km is an illustration of the order of magnitude, not an engineering project, and it will operate with adjustments for the same Baltic and Arctic winds as the NATO network. But even as an illustration, it demonstrates that the task is not solved by another S-400 regiment, but by a different class of systems.
NIF's deal with Kelluu is an indication that the alliance's pipeline is in motion. The Finnish 12-meter satellite itself doesn't change the balance of power on the eastern flank. What's important is the pipeline: the process by which the alliance brought the idea to Series A status within the timeframe of a single Russian conceptual design.
The Russian defense industry is capable of building a hydrogen airship of this class: the expertise and components are all in place, even taking into account the sanctions-related shortages. It hasn't been built for another reason: Russia lacks a channel between the practical idea and a serial order, similar to the NIF.
Information