The $117 billion between the satellite and the decision

In June 2026 the World Economic Forum and Deloitte published Earth Observation: Unlocking the $700 Billion Opportunity. The headline number is big, but the useful number is buried a few pages in — and it happens to be the exact problem we work on.

Where the value is actually stuck

The report estimates Earth observation (EO) could add $703 billion to global GDP each year. About $440 billion of that is already captured. The remaining $263 billion is unrealized — and, crucially, not because the science is missing or the use cases are speculative. It’s unrealized because insight isn’t reaching the systems where decisions get made.

The report splits that $263 billion across three bottlenecks — and the largest by far is the one we work on:

The largest slice — nearly half — is processing: the WEF’s term for the distance between a raw observation and a decision-ready output. Not launching more satellites. Not training a better model. The gap between data that exists and data an organization can act on.

Where the value concentrates

The unrealized value isn’t spread evenly. Seven use cases account for more than 99% of it — and the two largest, precision agriculture and risk & vulnerability analysis, are both processing-constrained in the report’s own breakdown:

It’s a data-architecture problem, not a space problem

Read the report’s description of why processing is stuck and it stops sounding like a space-industry issue and starts sounding like a data-engineering one. The frictions it names: weak interoperability, poor fusion across sources, value “lost at the handoff points,” and processed outputs delivered in — in the report’s words — “formats that enterprise systems cannot ingest” (WEF/Deloitte, 2026).

Anyone who has worked with EO data recognizes the shape of this. The archive is petabytes of multi-dimensional arrays — latitude × longitude × time × band — sitting in object storage as Zarr or NetCDF. To ask a question of it, the default playbook is: pick a subset, copy it out, convert the format, reshape it, load it into a warehouse or a bespoke Python job, and then run the analysis. Every team does it again. Every tool does it again. Every time the upstream data refreshes, the copy goes stale and the pipeline runs again.

That copy-and-convert tax is the processing friction the WEF is pricing at $117 billion. We wrote about its economics in What it costs to copy the planet’s data — the egress, the duplicated storage, the stale copies, the standing pipelines someone has to babysit.

The fix the report is already describing

Here’s the part that made us sit up. The report frames processing as a maturity ladder — from Initial (hand calibration) to Integrated (automated, analysis-ready pipelines) to Native (decision-ready analytic products) — and the enablers it names for climbing it read like a design spec for query-in-place:

• analysis-ready data in open geospatial formats on cloud object storage;
• cloud-native architectures with interoperable, API-first delivery that reduce dependence on any single platform;
• multi-source fusion — EO combined with weather, sensor, claims and supply-chain data;
• and, on the deployment side, interactive querying so non-specialists can ask questions in familiar terms.

Put those together and you’ve described a SQL engine that runs directly against the array data where it already lives. That’s exactly what we built zarr-datafusion to do: register a cloud-hosted Zarr or NetCDF archive as an external table and run SQL over it in place — no download, no conversion, no warehouse load. Multi-source fusion stops being a pipeline and becomes a JOIN. And “interactive querying” is just SQL (or an LLM writing the SQL) over the same store.

We put a hard number on it recently: we reproduced NOAA’s official El Niño index to 0.001 °C by pointing SQL straight at the source NetCDF via VirtualiZarr — no copy at any step. That’s the processing gap closing, on one worked example. The architecture behind it is the same one the WEF is asking the industry to adopt.

Where this leaves you

The $117 billion isn’t waiting on a breakthrough. It’s waiting on the boring, decisive work of making EO archives queryable where they sit — so an insurer sizing catastrophe risk, an agronomist scoring a field, or an energy planner reading tomorrow’s wind can get an answer without first standing up a data pipeline.

That’s the layer we build: not another data source, not another dashboard — the processing fabric that turns a cloud-hosted array into a decision, in place.

If your EO value is trapped between the satellite and the decision, that’s the gap we close. Reach us at [email protected].

---

Figures and framing from the World Economic Forum & Deloitte, Earth Observation: Unlocking the $700 Billion Opportunity (June 2026). This post is our independent commentary and does not imply endorsement.