Edge Computing in Space: How Novi Is Rewriting the Economics of Earth Observation
Earth observation is at a breaking point. We have more satellites than ever, more sensors in orbit, and more raw data streaming to the ground every day. Yet most of that data is never used. And the cost of collecting, storing, and analyzing it has kept the industry locked behind government budgets, exclusive contracts, and slow workflows.
But a major shift is underway. A shift that looks a lot like the moment smartphones replaced feature phones. A shift driven by edge computing in space.
In my conversation with Michael Bartholomew of Novi, we explored how satellites are evolving from passive collectors into smart devices with onboard compute, and why this change could reduce costs by orders of magnitude, unlock new markets, and reshape how we work with Earth observation.
Why the Current Model Is Breaking
Today’s satellites behave like giant cameras in orbit. They constantly capture data, transmit massive raw files to the ground, and rely on expensive earthbound systems to store, curate, and process everything.
Only a fraction of that data is ever used. Yet everyone pays for all of it.
This model creates three major problems:
1. Massive Waste
We store terabytes of data that only partially drives meaningful insights.
2. High Cost
Downlinking, storing, and processing raw imagery requires heavy infrastructure and enormous budgets.
3. Limited Access
Most organizations simply cannot afford Earth observation at meaningful scale.
This is why today’s EO market is dominated by governments and large enterprises, not the broader set of industries that could benefit from spatial insight.
Edge Computing in Space Changes the Equation
Novi isn’t trying to solve this from the ground up. They’re solving it in orbit.
Instead of sending all imagery to Earth for processing, Novi satellites run inference directly on the satellite itself. Only insights, detections, and clipped results come back down.
This turns gigabytes into kilobytes.
Michael shared one example that makes the economics crystal clear:
- Counting ships in a harbor using traditional EO: $120,000–$150,000 per year
- Doing the same task with Novi’s edge-processed insights: under $1,000
That’s not a 20 percent improvement. That’s a two-order-of-magnitude leap.
When compute lives next to the sensor, everything changes:
- Lower latency
- Lower cost
- Higher tasking frequency
- More organizations able to participate
- Faster, more precise insights
And for EO professionals, this flips the entire workflow. Instead of downloading full scenes or multi-band imagery, you ask a question and get the answer.
Why This Shift Is Happening Now
Edge computing on satellites isn’t a new idea, but it only becomes practical when several trends collide:
Miniaturized, low-power compute
Hardware is now small enough and efficient enough to run serious inference models in orbit.
Proven sensor packages
High-quality multispectral, hyperspectral, and RF sensors are well-tested and increasingly affordable.
Launch economics transformed
SpaceX and other providers have driven launch costs down dramatically.
Software deployable to orbit
Novi supports an SDK model, meaning developers can ship their models to satellites the same way they deploy cloud functions.
As Michael put it, satellites are becoming the orbital equivalent of smartphones: sensor suites with onboard compute, programmable by developers, creating applications we can’t even imagine yet.
From DoD-Only to a True Dual-Use Platform
Novi’s first chapter was in national security, where latency can be the difference between success and disaster. They built and deployed hardware, compute, and mission pipelines for the Department of Defense.
But while solving latency, they realized they had also solved cost. And cost is what unlocks the commercial market.
That insight pushed Novi toward a dual-use constellation designed from day one to serve both government and commercial customers. The focus is simple:
- Multi-sensor satellites
- Onboard compute
- Open access
- Developer-friendly tooling
- Liberal data rights
That last point is a big one.
Most EO providers heavily restrict what you can do with the data you purchase. Even if you can afford the imagery, you might not be able to resell, redistribute, or monetize it. Novi is breaking that model by giving users ownership rights over the data they task, allowing them to build businesses, models, and analytics workflows freely.
Precision vs Scale: Where EO and AI Intersect
Today’s AI landscape pushes toward foundational models trained on massive volumes of multi-sensor data. Earth observation is essential for that future.
But there are two distinct layers:
- Foundational models, which require enormous amounts of raw data and compute on Earth.
- Insight-driven EO, which requires fast, low-cost, high-frequency answers.
Novi lives firmly in layer two, but their outputs can feed layer one. Insights from orbit can inform, clean, supplement, or guide the models trained on the ground.
These are overlapping, not competing, swim lanes. And both are necessary.
Industries That Benefit First
The moment you cut cost and latency, entirely new classes of organizations can make use of Earth observation.
Here are the sectors Michael expects to move early:
- Wildfire monitoring and emergency response
- Forestry and land management
- Maritime monitoring and dark fleet detection
- Oil and gas infrastructure
- Environmental compliance
- Precision agriculture
- Mining and exploration
- Logistics and supply chain monitoring
- Hedge funds and investment research
And over time, entirely new use cases will emerge — just as smartphones created apps we could never have predicted in 2007.
One example stuck with me: a model that detects and tracks elephants. With cheaper, faster orbital compute, ecological and conservation use cases like this could explode.
What Using Novi Will Actually Look Like
Novi’s platform is designed for three user groups:
1. Enterprises and Government Agencies
- Private workspace
- SDK for deploying models to orbit
- Task satellites directly
- Receive raw, clipped, or insight-level outputs
- API-driven pipelines
2. Smaller Commercial Users
- Point-and-click ordering
- Lower-cost clipped scenes
- Off-the-shelf insights
- Ability to deploy their own lightweight models
3. Developers
- A sandbox environment
- Full SDK access
- Ability to publish applications to a marketplace
- Monetization tools to sell insight-generating models
This is the App Store moment for Earth observation. And it appears to be arriving much faster than expected.
The Timeline
Novi’s first two multi-sensor satellites — equipped with RGB, hyperspectral, and potentially RF — launch within the next six months on SpaceX Transporter 16 and 17.
Once those are operational, the constellation will grow aggressively, aiming for true global persistence.
The Big Idea: Making Space Boring (In the Best Way Possible)
The most important part of this entire conversation wasn’t a sensor spec or a compute benchmark.
It was the acknowledgment that we do not yet know the most important future use cases. Just like no one in 2007 predicted ridesharing, mobile banking, or TikTok, we are standing at the beginning of an era where space becomes programmable, accessible, and low-cost.
Lower cost means more innovation.
More innovation means more use cases.
And more use cases mean a fundamentally different role for Earth observation in the modern world.
Novi isn’t just shrinking satellites. They’re shrinking the barrier between a question and an answer.
