At INTERGEO 2026, Alberto Perez Cassinelli spoke with Geoawesome about combining Earth observation, onboard processing and satellite communications. The conversation began with the case for another constellation and ended with a harder question: what makes the information it delivers trustworthy?

Why another satellite constellation?

That was the question I put to Alberto Perez Cassinelli, VP of Data at Open Cosmos, when we sat down at INTERGEO. I wanted to understand what the company was proposing beyond another source of imagery. For the people using Earth observation, the important question is what those observations allow them to do.

Perez Cassinelli’s answer brought together two businesses that might otherwise be discussed separately: observing the Earth and connecting it. Open Cosmos, he explained, is developing Earth observation and telecommunications constellations within a common architecture. The intention is to shorten the path between collecting a measurement in space and getting useful information into someone’s hands.

“What we are doing is creating the ultimate machine for understanding and connecting our world,” he said. It is an expansive description, but the practical idea behind it is worth exploring: what if a satellite could do more of the analytical work itself, then pass the result through a communications network rather than treating collection, processing and delivery as separate problems?

That question became the thread running through our conversation. It connected the company’s near-real-time ambitions to disaster response, its shared infrastructure model to national control, and its plans for AI-ready data to a discussion about provenance. The more we talked about moving information quickly, the more important it became to ask what happens to that information along the way.

Bringing observation and communication together

When Perez Cassinelli described the relationship between imaging and telecommunications, I asked what it would mean in practice. Was the idea to relay information through other satellites rather than wait for a ground station pass?

His answer went beyond transmission. “We are going to provide not only imaging capacity but more processing capacity,” he explained. Customers would be able to process products in space and use the telecommunications constellation to deliver the results.

In that model, processing is part of the observation service, not simply something that happens after an image reaches a customer. Perez Cassinelli described DataCosmos as the access point for collecting new imagery, finding archive data and, as the onboard capabilities develop, running customers’ own analytics in orbit.

OpenConstellation 1.0 gives that architecture a specific ambition: actionable intelligence in as little as 30 minutes, combining rapid tasking, onboard AI processing and communications through ConnectedCosmos and partner networks, according to the company’s announcement. The first two satellites, Posidònia and Hyperion GR-1, launched on 7 July 2026, according to the launch announcement.

In the interview, Perez Cassinelli described a planned 15-satellite constellation, with four to five revisits a day and collection capacity of approximately three million square kilometres daily. Those are the intended capabilities of the constellation, rather than a description of what two spacecraft alone can provide.

The 30-minute figure also needs to be understood in context. The DataCosmos product page distinguishes 30-minute insights delivered through inter-satellite links from four-hour rush delivery and 24-hour standard delivery. It is the company’s best-case proposition, not a promise that any location can be observed and analysed within half an hour of a request.

For a user, however, the underlying ambition is easy to understand. Reducing the time spent moving and processing data could leave more time to act on what it reveals. Disaster response is where that proposition becomes particularly tangible.

When the useful product is an answer

I asked Perez Cassinelli where he saw the strongest demand. Defence remained important, he said, but his answer quickly broadened to civil government: wildfires, floods and earthquakes, alongside maritime monitoring.

Consider the questions a response team might need to ask after a flood. Which areas appear to be inundated? Has the situation changed since the last observation? Where should teams seek further confirmation before committing resources? These are questions about a developing situation, not requests for imagery for its own sake.

The same distinction applies to a wildfire, earthquake or landslide. An image can be part of the evidence, but its operational value depends on what can be established from it and how quickly that finding reaches the people responsible. A rapid but ambiguous result may still require interpretation; a well-understood result delivered too late may miss the decision it was meant to support.

Open Cosmos’s government missions give this ambition a concrete setting. The Posidònia mission is intended to support the Balearic Islands in responding to fires, storms and floods, while Hyperion GR-1 lists flood monitoring and disaster assessment among its applications. These are intended uses, rather than evidence that the complete 30-minute service has already been demonstrated across those hazards.

There is also earlier work behind the onboard-processing approach: on ESA’s Φsat-2 mission, built and operated by Open Cosmos, the Sat2Map application was designed to generate maps for teams responding to disasters including floods and earthquakes, according to the mission announcement. The relevance is the approach itself: undertaking analysis close to where observations are collected, then transmitting information selected for its usefulness.

When I suggested that this could mean sending results rather than every pixel, Perez Cassinelli put it simply: “We can sell the pixels, but we also sell the results.” Customers, he added, could use the infrastructure to deliver those results to their own end users.

That does not make imagery unimportant. It changes what the service is being asked to deliver. The customer may want to inspect an image, run a specialised analysis or receive a result inside an existing workflow. The proposition is to support those different needs without making every customer assemble the entire chain themselves.

Who gets to decide where the satellites look?

For a government, a faster service is only part of the equation. It also matters whether the required collection capacity will be available when needed, and who controls it. That brought us back to the business model behind OpenConstellation.

Perez Cassinelli described a shared system built around anchor customers. A country commissions a mission and controls collection over its area of interest. Outside that area, Open Cosmos uses available capacity to serve other customers; in return, the anchor customer gains access to capacity from the wider constellation.

He described the intended benefit as doubling or tripling the customer’s collection capacity within its area of interest. “We deliver sovereign solutions with a lower CAPEX,” he said. “We lower the entry barrier for many countries and governments.”

When I asked whether this additional capacity meant buying existing images or commissioning new observations, he was clear: it was about collecting new imagery through tasking. The distinction matters because access to an archive and the ability to request a fresh observation answer different needs.

There is an industrial dimension, too. Perez Cassinelli emphasised local integration and the effort to involve local suppliers. In his account, the proposition is not simply access to a satellite, but participation in the capability behind it.

This helps explain how the commercial model connects to the technical ambition. A shared constellation is intended to offer more collection opportunities than an individual national asset, while preserving defined customer control. For governments, the detail of those arrangements will matter as much as the headline performance: faster delivery is useful only if they can obtain the observation they need.

The next user may be an AI system

From there, our conversation moved to another kind of customer. I asked Perez Cassinelli about foundation models, and he described plans to develop an AI-ready embedding layer on top of OpenConstellation 1.0 for AI agents. He presented it as work ahead, not an already available product.

His broader expectation was clear. “We see in the future a big demand coming from AI-based systems,” he said. “We believe that Earth observation is a critical data for knowing our physical world.”

For me, the important connection is between that demand and the earlier discussion about new collections. A model can work with the observations it has, but an older image cannot establish what happened after it was acquired. Historical data remains valuable as a baseline; a current observation allows a system to test whether that baseline still describes the situation.

That is also why the discussion extends to digital twins, which Open Cosmos identifies among the integration targets for its data services. Consider a twin intended to support decisions about a changing landscape. Its usefulness should depend not just on how convincingly it represents the world, but on which conditions have actually been observed, when they were measured and how confidently they can be interpreted.

Perez Cassinelli’s sensor strategy fits that ambition. Rather than prioritising a single type of imagery, he described combining multispectral, shortwave infrared and hyperspectral observations into a common product. His emphasis was on the combination of spatial and spectral information and the applications it could support.

The attraction is more than faster refreshes. It is the possibility of providing different measurements of the same changing world in a form that analytical systems can use. But making information easier for machines to consume also makes its history more important: an output should not become harder to question simply because it arrives ready for an AI workflow.

Faster information still needs a history

That was the point at which I pressed Perez Cassinelli on trust. If a product is passed between providers, processed into another form and used as input to an analytical system, how does the eventual user know what they are looking at?

He described metadata, unique identifiers and checksums as elements of Open Cosmos’s approach. I challenged the distinction between identifying a product and establishing that its pixels had not been altered. He acknowledged that pixels could be changed and argued for comparing the data used in analytics against the original.

I then raised a complication: processing intended to improve an image, such as removing cloud cover, also changes it. The question is therefore not simply whether a file has changed, but what was changed, by whom and for what purpose. We were no longer talking only about the source of an image; we were talking about the history of a result.

Perez Cassinelli returned to traceability across the supply chain. “I don’t think there is an ultimate solution yet for this,” he said. It was a useful acknowledgement of the difficulty, rather than a claim that an identifier alone resolves it.

We had started with why the industry needed another constellation. By the end, the more interesting question was what users should expect from one. Open Cosmos is proposing to connect collection, analysis and delivery more closely, with a business model designed to share capacity without eliminating national control.

The promise is not merely that information will arrive sooner. It is that the information will arrive in a form someone can use. As more of that interpretation happens in orbit or inside AI systems, the ability to trace the result back to its observations becomes part of the service itself.

Selling results rather than only pixels is a compelling direction. The strongest version of that proposition will give users both an answer they can act on and enough visibility to understand why they should trust it.

Interview excerpts: inside the operating model

The following excerpts add detail on operations, local industry and customers. Questions and answers have been condensed and edited for clarity; researched context has not been inserted into Alberto’s responses.

Do you hand a satellite over to the customer, or do you continue operating it?

Alberto Perez Cassinelli: We have different models. One is called OpenOrbit. With that model, we deliver the mission turnkey to the customer, so typically we transfer operations.

With OpenConstellation, the mutualised model, we operate the spacecraft. Governments control the tasking and the capacity over their area of interest.

How important is working with local industry to your approach?

Our approach is mainly this 2x–3x model with local integration. That is important. We work together with local industry and integrate the satellites locally. Typically, we put together a factory where we deliver the solutions, and we make a big effort to integrate local providers.

Are your customers mainly in Europe?

We have customers worldwide, mainly in Europe, but from Japan to Argentina. Regarding factories and offices, our main capacity is in Europe. We have factories at our headquarters in the UK, and also in Spain, Portugal and Greece.

Do you intend to prioritise a particular application?

Real time for emergencies, defence and intelligence is one of our competitive edges. The other is our combination of spatial resolution and spectral information in a broader product that can unlock applications for agriculture, for instance.

What does DataCosmos allow customers to do?

It is a space intelligence platform, the access point to the constellation for tasking satellites and accessing the archive. We deliver data services through the platform.

Users can upload their own analytics to the satellite, run products onboard with edge computing, and deliver the results using the constellation’s inter-satellite links. Customers can use our infrastructure to deliver results to their final users.