Europe-UAE Orbital Data Center Needs a Government Buyer
Macron used the Paris space summit to put a number on the Altair satellite program: $1 billion to grow the constellation from 10 satellites to 50, each carrying Mistral AI models. At about $20 million a satellite, the economics work only if defense and intelligence agencies sign as the first customers.
Macron used the International Space Summit in Paris to put a number on something European and Gulf officials had been discussing since early September: a $1 billion program, led by Abu Dhabi's Marlan Space and San Francisco-based Loft Orbital, that will grow the Altair Earth observation constellation from 10 satellites to 50, as TechRadar reported. The satellites are being built by Orbitworks in Abu Dhabi and will run Mistral AI's models onboard, under the project name Altair Next Gen. None of the first ten satellites, the ones already in production, has launched. Full deployment is planned over five years.
The term "data center" can mislead. A data center rents general purpose compute; Altair Next Gen sells inference next to the sensor: radar and optical instruments feed images and signals into onboard AI models that interpret them in orbit, so a customer gets an alert in seconds instead of raw imagery that has to be downlinked and processed on the ground, a step that can take hours. Customers do buy this kind of service, but it is narrower than "data center" implies. Its economics work only if buyers pay a premium for speed rather than cheap compute.
At $1 billion for 50 satellites, the program runs about $20 million per satellite before ground systems and operating costs. Commercial Earth observation has never produced enough recurring revenue on its own to support spending at that level. The customers who have reliably paid for space based surveillance are governments, particularly defense and intelligence agencies, and that is who this constellation is built for. The consortium reflects that aim: Emirati capital through Marlan, a French model maker in Mistral, and an announcement from the French president himself. Agencies that want surveillance capability independent of American satellite and cloud providers are exactly the buyers who would fund a European or Gulf alternative, and exactly the ones likely to place the first large order.
Altair Next Gen has a different business model from the other orbital computing programs announced this year. SpaceX and Blue Origin have filed plans for general purpose data centers in orbit, betting they can eventually beat ground facilities on cost as launch prices keep falling. Google's Project Suncatcher puts its own chips in orbit for the same reason, cheaper compute at scale. Both bets need space based compute to underprice terrestrial data centers, an argument that has not yet been proven at any scale. Altair Next Gen does not need to win that argument. It sells a faster alert on a fire, a troop movement or a ship that has gone dark, to a customer who already values speed over price.
Loft Orbital's chief operating officer, Alex Greenberg, described the plan as building "a 50-satellite, AI-enabled, multi-modal constellation," with more partners still to be added. Part of the pitch, an application store where customers pick which open source model runs against their sensor feed, points toward something more commercial: developers renting orbital inference the way they rent cloud compute today. The desk doubts this becomes the growth driver. Commercial demand for Earth observation imagery has stayed thin for as long as the industry has tried to sell it, held back by the small number of buyers who need frequent, high resolution coverage badly enough to pay for it. Sovereign customers, the security services and militaries that already budget for satellite reconnaissance, are the ones who have reliably funded capacity like this, and they are the ones this consortium is positioned to sell to first.
Mistral benefits mainly through distribution, rather than direct revenue: a French government backed display of its open models running on French and Emirati government infrastructure, at a moment when Mistral is competing for enterprise and sovereign AI contracts against much larger American labs. It costs Mistral little to supply models to a fleet of 50 satellites, and it buys the company political goodwill in Paris and Abu Dhabi that is harder to price than the satellites themselves. The satellite operators carry the capital risk; Mistral carries almost none.
The program remains unproven. Its first ten satellites, the base for the 50-satellite expansion, have not flown. The program has effectively been announced twice, first in outline on September 9 and now with a dollar figure at the Paris summit, which is a normal pace for a diplomatically brokered initiative but also means two announcements have passed without a single satellite reaching orbit. A five year deployment timeline is long enough that the technology, the competitive field and the customer list could all look different by the time it completes. Loft Orbital and Marlan already have government relationships, through Loft's hosted payload business and Marlan's position in the UAE's space sector, so the consortium is not starting cold.
A signed contract committing government funding will show whether the plan is working, well before a product demo or another partner joins the consortium. The desk expects the first anchor customer, most likely a French or Emirati government agency, to commit before any commercial revenue appears on Altair Next Gen's books. Until the first of the original ten satellites reaches orbit, the fifty satellite figure is a funding commitment rather than a fleet, and with a five year buildout, meaningful revenue, government or commercial, is unlikely to show up before 2028.