2026.10.01 | Substack | Aerospace | BlackSky Physical AI Thesis Revisited

There was a BlackSky announcement yesterday morning that looked pretty small at first, but the more I dug into what they are actually doing, the more it brought me back to why I originally put $BKSY in my Physical AI portfolio.

BKSY Deep Dive

Gaetano

·

Jun 14

Read full story

Robinhood is rolling out something called Agent Apps, which basically lets its AI agent connect to specialized outside data sources while you are researching a stock.

One of those new apps comes from SkyFi, a software platform that makes commercial satellite imagery easier to access. SkyFi doesn’t own the satellites itself. It connects to companies that do, pulls their imagery into one platform, adds analytics on top, then makes all of that available through software. BlackSky is going to be the first satellite imagery provider powering this new Robinhood product.

The example that immediately stood out to me was AI data centers. SkyFi is launching an AI Buildout Monitor that will use satellite observations to track the physical progress of hyperscale campuses over time. It can look at things like how much land has been cleared, foundations going in, cooling infrastructure showing up and the electrical build around the site. I personally don’t spend my days tracking individual data-center campuses from satellite imagery, but I know plenty of people invested in names like NBIS, WYFI, IREN etc. that absolutely do. A huge part of the debate around those stocks is whether these enormous announced buildouts are actually happening on schedule, how quickly capacity is coming online and whether management is executing against what they told investors.

Right now people dig through permits, utility filings, construction updates, local news, drone footage and posts on X trying to piece that together. Now imagine being able to ask your Robinhood agent how much physical progress has been made at a specific campus over the last few months and have it pull satellite observations, compare them and show you what actually changed. You still have to interpret that carefully because seeing more concrete or cooling equipment doesn’t tell you exactly when GPUs are going live, but it is another piece of physical evidence that can be layered into the research.

That is cool on its own. What interested me more was what it says about BlackSky.

When I originally bought BKSY, I put it in the data layer of Physical AI because my thinking was that AI systems operating in the real world would eventually need a broader understanding of what is happening in that world. I actually think I was describing that too vaguely at the time. A robot doesn’t need BlackSky to tell it where the wall in front of it is obviously. The robot has cameras and perception. A drone can see the power line it is inspecting. And so on. BlackSky starts becoming interesting one level above those machines.

Imagine a utility has thousands of substations and transmission assets spread across a state, along with a fleet of inspection drones. A hurricane comes through. Once a drone gets to a substation, its own cameras can inspect the equipment far better than a satellite ever could. The problem is figuring out which locations need to be inspected first. That is where I can see something like BlackSky fitting. Satellite observations cover a huge area, software flags locations where something appears to have changed, the utility’s planning system decides those locations deserve attention, then the drones get sent there and local perception takes over.

The same idea can work in defense. An autonomous drone can have incredible perception and still only know what is happening within the area its sensors can see. Something above that fleet needs to decide where the drones should go in the first place. Did vehicle activity suddenly increase around one location? Did an airfield change? Did equipment appear somewhere it wasn’t yesterday? Did a bridge get damaged?

BlackSky already collects imagery and runs AI against it to detect objects and changes. Over time, I can imagine more of that information feeding directly into the software coordinating the physical assets instead of stopping at a human analyst looking at an image.

That is why I think today’s Robinhood announcement is a fun little glimpse into the plumbing.

An AI agent has a question about something happening in the physical world, it reaches outside of its own model, pulls satellite information, gets that information analyzed and brings the answer back into its workflow. Today the workflow is an investor trying to figure out whether an AI data center is getting built. There is nothing particularly “Physical AI” about that use case itself. The interesting part is that software can ask the physical world a question and get a machine-readable answer back.

Change who is asking the question and things start getting more interesting.

Maybe it is utility software deciding where inspection drones should go after a storm. Maybe it is logistics software looking across ports and distribution hubs before reallocating trucks. Maybe it is defense software coordinating autonomous aircraft across a much larger area than any one aircraft can see.

BlackSky also has another interesting angle here through Project Orbion, where its imagery is being used as part of an effort to build and update digital environments for training autonomous systems. That could eventually become another piece of this. Roads change, buildings go up, infrastructure gets added, terrain changes and equipment moves. If autonomous systems are going to spend more time training inside simulations before operating in the real world, keeping those simulated environments closer to reality becomes useful too.

There’s another part of this announcement I like too. BlackSky is still heavily associated with defense and government intelligence, which has been a major part of my thesis from the beginning. This is a very different use case. Now its imagery is getting packaged into a consumer investing platform so someone can track something as commercial as data-center construction. That starts showing me how the same infrastructure can branch into other markets. Construction monitoring, logistics, energy infrastructure, mining, insurance, agriculture, supply chains. Anywhere physical activity changes over time and somebody is willing to pay to understand it, BlackSky potentially has something to sell.

I think that is important because the bigger BKSY story gets a lot more interesting if defense becomes the foundation rather than the ceiling. The government contracts help fund the constellation, prove the technology and create demand for high-cadence imagery. Then the same satellites can serve commercial customers through platforms like SkyFi. If that happens, you start getting more utilization out of the same physical infrastructure and more ways to monetize every satellite already in orbit.

So I want to update/sharpen my original BKSY thesis a little.

I used to describe BlackSky as the “see the world” layer of Physical AI. I think what I am actually looking for is whether BlackSky can become part of the wide-area intelligence layer sitting above physical machines, continuously giving software a better understanding of what is happening across areas those machines cannot see themselves.

That would take BKSY well beyond selling satellite pictures.

It starts looking more like a data source for the software coordinating what happens in the physical world.

And today we got a very small, very understandable example of how that information can start flowing directly into AI.

Subscribed

Financial Disclaimer: I own shares of BlackSky Technology ($BKSY). This article reflects my personal research, opinions, and investment process and is provided for educational purposes only. It is general market commentary rather than individualized financial, tax, or legal advice. I may add to, reduce, or exit any position discussed as new information, valuation, price action, portfolio construction, or risk considerations change. Please conduct your own due diligence and consider your own financial situation before making any investment decision.