1. The matter compiler in reverse

Neal Stephenson’s The Diamond Age describes matter compilers that turn information into physical objects. The desktop does the opposite: it turns photons into information, and then sells the information. The question is what form the information takes and what customers will pay for it.

Entry 234 defined the onboard processing pipeline. This entry defines the products that come out of it and the prices that make the desktop’s sensor commercially real.

2. Product tiers

The desktop should sell four tiers of product, each aimed at a different customer tolerance for cost, latency, and processing effort.

Scenes. A single recent image of a specified area, delivered as a web-ready visual product. This is the impulse-buy tier: a farmer checking a field, an insurer verifying a claim, a journalist illustrating a story.

Archive slices. Access to all historical imagery over a region for a period, delivered as a stack of geoTIFFs or via an API. This is the research and analytics tier.

Change feeds. A subscription that delivers a new image and a machine-readable change mask whenever the area is revisited. This is the operational monitoring tier.

Raw data. The level-1A radiometrically corrected frames, for customers who want to run their own algorithms. This is the expert tier and is priced higher because it consumes more downlink and storage.

3. Pricing anchors

Free Sentinel-2 data at 10 m resolution sets a floor of zero for basic multispectral coverage. Commercial providers at 0.3–0.5 m resolution task at hundreds of dollars per square kilometre. The desktop’s 3–5 m product sits in the gap.

The price must be low enough to beat high-resolution tasking for routine monitoring, and high enough to cover the desktop’s marginal cost of imaging, processing, and downlink.

A reasonable anchor:

  • Scenes: $5–20 per 100 km².
  • Archive slices: $0.50–2 per km² for the full historical stack.
  • Change feeds: $50–200 per month per 1,000 km².
  • Raw data: $20–50 per 100 km².

These are order-of-magnitude figures. The exact price depends on commitment length, area, and whether the customer needs priority tasking.

4. The subscription model

The most stable revenue comes from subscriptions rather than one-off scenes. A customer commits to a region and receives every usable image the desktop collects over it. The desktop gets predictable demand and can schedule imaging efficiently.

A base subscription should include:

  • All cloud-free images collected in the subscribed region.
  • Level-1C geoTIFFs in four bands.
  • A cloud-mask band.
  • API access to the archive for the subscribed region.

Premium tiers can add:

  • Higher priority in the imaging queue.
  • Off-nadir tasking outside the normal ground track.
  • Raw data access.
  • Custom-derived products such as NDVI time series.

5. Customer payload pricing

Customer sensors that plug into the attachment grid are priced differently. The desktop charges for:

  • Power: per watt averaged over the orbit.
  • Data: per gigabyte stored and per gigabyte downlinked.
  • Compute: per CPU-hour or GPU-hour used for processing.
  • Mechanical: a flat integration and qualification fee for the first unit.

This unbundling lets customers bring sensors that are heavier or thirstier than the baseline imager without distorting the desktop’s own economics.

6. The free tier

Some imagery should be free. A low-resolution browse product, delayed by a few days, can attract developers, researchers, and journalists. The free tier builds familiarity and creates demand for paid analytics.

The free tier also keeps the desktop honest. If the only reason customers pay is exclusivity, then the product is not actually better than free alternatives. If customers pay for recency, reliability, or derived analytics, the value is real.

7. Operational pricing rules

Several rules keep the pricing from becoming a maintenance burden:

  • Prices are fixed in a public price list, not negotiated per scene.
  • Subscriptions bill monthly in advance.
  • One-off scenes bill at purchase.
  • Custom tasking carries a surcharge and a minimum area.
  • Academic and humanitarian use gets a documented discount tier.

The desktop should not become a consultancy. The product must be standardised enough that customers can buy it through an API without talking to a sales team.

What this changes

  • The desktop sells scenes, archive slices, change feeds, and raw data as standard products.
  • Prices are anchored between free Sentinel-class data and expensive sub-metre tasking.
  • Subscriptions are the primary revenue model; one-off scenes are the entry point.
  • Customer payloads are priced by power, data, compute, and integration, not as bundled mystery fees.
  • A limited free browse tier builds demand and validates product value.
  • The next entry can close the sensor attachment arc and decide what attachment to define next.