TerraWatch Space

Become a Pro subscriber and get weekly exclusive market briefings! newsletter.terrawatchspace.com/earth-observat…

The latest edition of the @terrawatchspace Pro newsletter is out with curated news, analysis and insights on Earth observation and applications. 🎯 Energy Aspects acquires Kayrros: What this means and the trend of strategic acquisitions in EO 📈 ICEYE's Path to €1B in Revenues: How the SAR player averaged ~95% annual growth from 2020 to 2025

Planet has imposed a 14-day delay on satellite imagery across the Middle East. The company is US-headquartered and subject to US national security restrictions. The world has come to rely on commercial satellite imagery for independent journalism, conflict monitoring and humanitarian access. But that infrastructure is concentrated in a few companies, mostly American. When those companies restrict access - whether voluntarily or under pressure - there is no fallback. No international mechanism. No agreed principles. Just individual companies, under individual national jurisdictions, making unilateral decisions about who gets to see what globally. I am not saying Planet made the wrong call. These are hard decisions with real consequences for personnel on the ground. But right or wrong, the point is that there is literally no framework for making these calls, and no alternative how/when they are made. We have the International Atomic Energy Agency for nuclear energy and the World Meteorological Organization for weather. We have no governing body for Earth observation - arguably the most powerful diagnostic tool we have for understanding what is happening on the planet. A few months ago, I wrote an essay arguing for a Global Earth Observation Accord - a political framework for EO governance. This week is why. 94 countries are now pursuing sovereign EO programs. If there was ever a time to have this conversation, it is now.

"Edge computing" gets used loosely in Earth observation. It can mean anything from cloud masking to running foundation models in orbit. The differences matter because each step in the pipeline produces a different output, serves a different purpose, and has different implications for what reaches the ground. Here's how the full onboard processing pipeline actually works, step by step: 1. Acquisition: Sensor captures imagery (optical, SAR, hyperspectral, thermal, RF) 2. Onboard Calibration: Radiometric correction and noise removal. Sounds routine, but if this is done inconsistently, the output won't integrate with ground-processed data. This is one of the most underappreciated barriers to scaling edge. 3. Filtering: Discarding cloud-covered frames, empty terrain, irrelevant areas. This is where the first major data reduction happens. 4. Compression: Reducing data volume while preserving information. The most advanced approaches produce compressed representations that double as embeddings - meaning downstream AI models can use them directly without decompressing. 5. AI Inference: Detection, classification, feature extraction. This is where data becomes information. 6. Output: Results are either downlinked as analytics (alerts, coordinates) or sent via inter-satellite link to retask another satellite. What happens after step 6 is where it gets interesting. There are fundamentally different implementation pathways each with different trade-offs, different customers, and different business models. I break all of this down in the full deep dive for Premium subscribers.

It's been only a couple months since I published this essay on the inevitable journey of Earth observation from hype to invisibility. Glad to see the concept of EO as "invisible infrastructure" catching on. The only way EO becomes mainstream is when it becomes invisible. That remains my core thesis for the industry over the next decade. newsletter.terrawatchspace.com/the-earth-obse…

Kickstarted the "The Business of Earth Observation" course this week with live sessions with two cohorts of 36 people from 17 countries from public sector, investors, EO companies, non-profits and end-user organisations. This week we focused on how the EO sector is structured — from the origins in 1960s, the defense dependency of EO, the future of civilian EO programmes, the importance of policies in the EO market, and the emerging role of philanthropy. Week 1 Takeaway: “EO is a unique sector in which companies, governments and philanthropies compete but also depend on each other simultaneously.”

Early bird ticket registration is also open for those interested in attending: eosummit.com

A few more sponsorship slots are still open for those interested: cdn.prod.website-files.com/656727b97114e5…

Excited to announce the first group of sponsors for EO Summit 2026 in London coming up in a few months. I am delighted to welcome repeat sponsors like Pixxel, UP42, Esri, EarthDaily, Kuva Space and Attis as well as new ones Chloris Geospatial and Aistech Space. Building a conference focused on applications and end-users—not just satellite specs and tech demos—remains the hardest thing I do on a yearly basis. So, I'm very grateful for the continued commitment from these sponsors, without whom this event wouldn't be possible.

We have more sponsorship packages available. Check out the brochure for more details: cdn.prod.website-files.com/656727b97114e5…

Announcing the first round of sponsors for @EOSummitConf 2026: 💎 Platinum Sponsor: @PixxelSpace 🟡 Gold Sponsors: @UP42_ and @Esri ⚪ Silver Sponsors: @EarthDailyA, @KuvaSpace and @ChlorisGeo 🟤 Bronze Sponsors: @AistechSpace and Attis

Free preview below + full piece for Premium subscribers. newsletter.terrawatchspace.com/edge-computing…

New @terrawatchspace deep dive: "Edge Computing for Earth Observation - 2026 Edition." This is the most comprehensive piece I have published on edge computing. It covers: - What actually happens when a satellite processes data onboard - Three distinct implementation pathways for edge AI and why the differences matter - A commercial landscape map of 30+ companies - The architectural bet between edge processing and optical downlinks - An adoption framework: what has to be true for edge to scale beyond demos - Where this is heading — foundation models, orbital data centres and more

Foundation models are NOT a moat in Earth observation. They are increasingly available, licensable, and buildable. So what is? First, distribution matters - model builders need a path into real EO workflows. That is why Vantor integrating Google Earth AI makes sense: Google gets distribution into operational workflows and Vantor embeds state-of-the-art models on top of its own archive and customer pipelines. But there is a trap for EO companies: integrating a model and announcing "AI-based platform" doesn't automatically move you up the value chain. If what you sell is still access to data plus tools, you are still a data supplier and the margin stays thin. In other words, AI remains a feature, not the product. So, I think the real moat is productised outcomes: deliverables that customers can buy as a service, where you are accountable for quality, timeliness, and accuracy. Reliable enough that customers don't care which model powers them, and auditable enough that they trust it. TL;DR: The margin (and the value) enabled by foundation models lives in delivering AI-based outcomes, not in offering AI tools.

Become a Pro subscriber and get weekly exclusive market briefings! newsletter.terrawatchspace.com/earth-observat…

The latest edition of the @terrawatchspace Pro newsletter is out with curated news, analysis and insights on Earth observation and applications. In this edition: 💰 An assessment of EO satellite company funding by sensor - Optical vs SAR vs Hyperspectral vs Thermal vs RF ♟️ Breaking down Vantor's partnership with Google AI and EO foundation model strategies

I have run countless market sizing exercises to estimate the total addressable market (TAM) for Earth observation. But here is the truth: TAM doesn't work for EO. Why? When we try to size the "EO market," we either measure imagery sales (small) or we try to capture every decision EO touches (impossible). And EO is not a product category, it is an input - like electricity, like GPS, like weather data. It is embedded in insurance pricing, agricultural planning, logistical tracking, infrastructure monitoring, defence, climate policy etc. That is why TAM conversations about EO always feel off. Even though every sizing exercise we do at TerraWatch, we try to account for as much as possible (trust me, you cannot wing it). Somehow we always end up in the same place: either the client says the number is too small or I convince them that the value is too diffuse to measure. My conclusion? EO doesn’t have one TAM, it has dozens of problem-specific markets PS. Doesn't mean I am closing the door to these projects, always welcome!

2020s is turning out to be the decade of sovereign Earth observation. If (big if) Orbital Data Centers (ODCs) become viable, the 2030s could turn out to be the decade of ODCs. We are seeing a clear pattern recently: EO is becoming strategic infrastructure for national security, disaster response, and more, leading to EO investments like never before. So even if ODCs take longer than the mid-2030s timelines to mature, I think sovereign EO will extend to the full stack. Why route sovereign data through someone else's orbital compute infrastructure to process it there? The logical endgame is to control the stack end-to-end: data, processing, and delivery (along with launch and satellite manufacturing). In other words, sovereignty may shift from "who can collect the data" to "who can collect, process, store, and distribute it independently." We have seen this movie before with GNSS: duplication isn't waste, it is the point (redundancy becomes policy) PS. Personally there is another reason I hope ODCs work out: if they are efficient, they could reduce the growth in ground data centers and the associated energy, cooling, water use, and emissions. Sovereignty and sustainability might actually align here. PPS. Space is big, until debris makes it small.

Become a Pro subscriber and get weekly exclusive market briefings! newsletter.terrawatchspace.com/earth-observat…