1. Headline & intro
If the 2010s were about packing more satellites into orbit, the late 2020s may be about packing more kilowatts into each one. K2 Space’s upcoming launch of its Gravitas spacecraft is not just another NewSpace milestone; it is an early test of a far bigger idea: data centers in orbit, powered like miniature space stations but priced like commercial satellites. In this piece, we’ll look at what K2 is actually flying, why “high power” is suddenly the hottest metric in space, what this means for cloud providers and defence agencies, and where Europe fits in a future where compute is no longer bound to Earth.
2. The news in brief
According to TechCrunch, US startup K2 Space is preparing to launch its first large satellite, called Gravitas, on a SpaceX Falcon 9 as early as the end of March 2026. Founded in 2022 by former SpaceX engineers Karan and Neel Kunjur, K2 has raised around $425 million and was valued at about $3 billion in late 2025.
Gravitas weighs roughly two metric tons and unfolds a solar array with a span of around 40 meters. That hardware is designed to deliver about 20 kilowatts of electrical power to hosted payloads – an unusually high figure for a single commercial satellite. The mission will carry a dozen customer payloads, including experiments for the US Department of Defense, plus a high-power electric propulsion system. K2 plans up to eleven more satellites over the next two years and is already designing 100 kW-class platforms aimed at future constellations and in-orbit compute.
3. Why this matters
Gravitas is important because it shifts the competitive benchmark in orbit from size and price to power and available watts per kilogram. For years, the NewSpace story was about making satellites smaller and cheaper. K2 is arguing the opposite: that the next breakthrough is giving spacecraft vastly more energy to play with.
Who benefits first? Secure connectivity providers and defence agencies. More power means stronger signals, more bandwidth, and better resistance to jamming. A military communications payload sitting on 20 kW has much more room for directional beams, redundancy and on-board processing than the typical smallsat running on a couple of kilowatts at best.
Next in line are Earth observation and AI workloads. If you can run more inference directly in orbit – filtering, classifying and compressing images before they touch the ground – you cut down latency, transmission costs and ground infrastructure demand. Think of wildfire detection, maritime surveillance or battlefield intelligence where seconds matter: orbital pre-processing is extremely attractive.
The losers, at least on paper, are traditional prime contractors and the long tail of cubesat builders. K2 claims it can deliver a high-power platform at a price that undercuts bespoke GEO-class buses while out-muscling small satellites on capability. If that cost-performance curve is real, some legacy GEO and mid-power platforms will look stranded.
There is, however, a hidden challenge: launch economics. K2’s original pitch leaned heavily on ultra-cheap heavy-lift from rockets like Starship and New Glenn. Those vehicles are delayed and unproven at commercial scale. Gravitas has to make sense even when launched on today’s still-expensive rideshare pricing. Power in orbit is only valuable if the all-in cost per delivered kilowatt – including launch – beats alternatives.
4. The bigger picture
K2’s move plugs into three converging trends.
First, the militarisation and securitisation of low Earth orbit. The US is planning multibillion-dollar constellations for missile warning and defence; Europe is building IRIS² for secure connectivity; China is scaling its own LEO networks. All of these programmes demand more resilient, jam-resistant, flexible platforms – and that means more power.
Second, hyperscalers are flirting with the idea of an orbital extension of the cloud. Microsoft has already experimented with Azure-connected satellites; AWS has Ground Station and a long list of space clients; Google is pushing AI inference to every possible edge. For latency-sensitive or regulation-heavy workloads, orbit won’t replace Earth data centers. But for certain niches – low-latency financial links, defence contracts, deep-space missions, high-value scientific processing – high-power satellites could become specialised “edge regions” of the cloud.
Third, the satellite industry is maturing from “any payload that fits” to platform specialisation. Starlink and Kuiper are vertically integrated communications constellations. Traditional GEO players like Viasat build massive, single-purpose broadband buses. K2 is trying to carve out a horizontal layer: generic high-power “trucks” that others can load with sensors, radios and compute, including classified payloads.
Historically, we have seen a similar transition on Earth. Early data centers were built around whatever power and cooling they could get. Hyperscalers flipped the script: they design facilities from the grid connection backwards. K2 is effectively applying that hyperscaler logic in orbit – start from tens of kilowatts, then ask what interesting things customers can do with it.
The risk is that incumbents may respond faster than expected. Both established primes and constellation operators can decide to bulk up their own platforms. If Starlink V3 or a future IRIS² bus jumps in power, K2’s advantage narrows. The company is betting that its speed, in-house manufacturing (reportedly covering the majority of components), and focus on power density will outrun that response.
5. The European / regional angle
For Europe, K2’s launch is a wake-up call. The EU is already committing billions to IRIS² and national defence constellations, yet much of the underlying platform innovation – especially around high-power buses and orbital compute – is happening in the US.
European primes like Airbus, Thales Alenia and OHB certainly know how to build powerful spacecraft, but they mostly operate in the slower, procurement-driven world of GEO communications and institutional missions. On the startup side, Europe’s NewSpace champions (Isar Aerospace, Rocket Factory Augsburg, PLD Space, ICEYE and others) focus more on launch and sensing than on power-heavy “orbital cloud” platforms.
Regulation also complicates the picture. The EU’s AI Act, GDPR, the upcoming space law initiative and the Digital Services Act all intersect awkwardly with the idea of processing sensitive data in orbit. If a surveillance or AI inference workload runs on a US-built satellite carrying defence payloads, who has jurisdiction? Which export controls apply? Where does data residency sit when the server is in low Earth orbit?
For Slovenia, Germany, Croatia and other smaller EU states, this is not an abstract debate. They will consume secure connectivity, Earth observation and defence data from these constellations, and their own startups will either plug into a US-dominated orbital cloud or help build a European alternative. There is an opportunity for EU-backed “green” orbital compute that emphasises debris mitigation, transparency and data sovereignty – but only if policymakers and investors move quickly.
6. Looking ahead
The next 24–36 months will tell us whether K2’s thesis holds.
Technically, the Gravitas mission must prove three things: reliable deployment and power generation at scale; stable hosting of multiple third-party payloads; and safe, efficient operation of a very high-power electric propulsion system. Any serious anomaly will not just hurt K2; it will be ammunition for those arguing that the industry should stick with smaller, simpler buses.
Commercially, watch for who signs the first multi-satellite contracts. If early follow-ons come from the US Department of Defense and intelligence community, K2’s fleet may lean heavily dual-use and classified. If, instead, the company can announce deals with a hyperscaler or a major commercial operator, the orbital cloud narrative gains real credibility.
On the regulatory side, expect growing pressure for stricter debris mitigation and traffic management for large, high-power platforms that can climb between orbits. European institutions, in particular, are unlikely to accept a “move fast and break things” approach in LEO.
One under-discussed risk is concentration. If orbital compute ends up clustered on a small number of very powerful US-owned buses, we’ve simply recreated today’s cloud centralisation above our heads – with even less transparency. That should concern European regulators and defence planners.
Still, the opportunity is significant. If K2 and its peers succeed, by the early 2030s we could see dedicated "space regions" in cloud consoles, AI models pre-processing sensor data in orbit, and a new class of startups building software for an environment where power is plentiful but every gram is still expensive.
7. The bottom line
K2’s Gravitas mission is less about one satellite and more about a new layer of infrastructure: power-rich orbital platforms that look a lot like the first racks of a future space-based cloud. If Europe wants a say in how that cloud is built – and whose rules apply – it cannot afford to treat this as just another American launch. The question for policymakers, investors and founders is simple: will the orbital cloud be something we merely buy capacity on, or something we help design and govern from the start?
