SpaceX and Google both said no. Now his software is riding a SpaceX rocket next to Google’s

We have reached the point in the AI hype cycle where terrestrial data centers are starting to look like a bottleneck. Between land use permits, cooling requirements, and the sheer thirst for megawatts, building on the ground is getting expensive. The logical next step, at least according to a growing cohort of founders, is to move the compute to where the cooling is free and the real estate is infinite.

The Silicon Valley of the Stars

A new startup called Satlyt just pulled in $8 million to solve a problem most of us haven't even considered yet: how do you run heavy AI models on hardware that is moving at 17,000 miles per hour? The founder, a veteran of both Google and SpaceX, is effectively trying to build the Android of space. It is a classic move. When you see a dominant player building a closed, vertical garden, you build the open-source alternative for everyone else.

Currently, the space industry looks a lot like the early days of the smartphone. SpaceX is the Apple of the sky. They build the hardware, they write the software, and they control the ecosystem. If you want to play in their sandbox, you play by their rules. Satlyt is betting that the rest of the satellite manufacturers—the Boeings and Lockheeds and smaller hardware startups—don't want to be locked into a proprietary stack. They want a flexible operating system that allows them to process data in orbit rather than just beaming raw images back to Earth.

Why Builders Should Care About Latency

If you are building an AI application today, you are likely relying on AWS or Azure. You send a request, the server processes it, and you get a result. In the world of satellite imagery and global monitoring, this process is broken. Currently, a satellite takes a photo, waits until it passes over a ground station, dumps a massive file, and then a server on the ground processes it. By the time you get the insight, the event you were tracking might be over.

Moving the compute to the satellite—what the industry calls edge computing in orbit—changes the game. Instead of sending a high-resolution photo of a forest fire, the satellite runs a lightweight model, identifies the fire, and sends a tiny text alert. That is the difference between a thirty-minute delay and a three-second delay. For builders, this opens up a whole new class of real-time applications in logistics, climate tech, and defense.

The Challenge of Vertical Integration

The biggest hurdle for Satlyt isn't the technology; it is the gatekeepers. SpaceX and Google both reportedly passed on this vision internally. That is usually a sign that a project is either too early or it threatens a lucrative status quo. For SpaceX, keeping their software proprietary is a competitive moat. For Google, the cloud is a terrestrial business. A startup coming in to decentralize that compute is a threat to the margin.

However, we have seen this play out before. The history of technology is a constant swing between vertical integration and horizontal fragmentation. We are currently in the integrated phase of space tech. Companies like Satlyt represent the beginning of the fragmentation phase, where specialized software layers start to emerge that work across different hardware vendors. As a founder, you want to be looking at where these layers are forming.

The Reality Check: Space is Hard

I have a healthy skepticism for anything that combines the two most expensive buzzwords of the decade: Space and AI. The hardware constraints in orbit are brutal. You are dealing with radiation that flips bits in your memory, extreme temperature swings, and a very limited power budget. You cannot just throw an H100 into a CubeSat and hope for the best.

The $8 million seed round is a drop in the bucket compared to what it takes to actually scale a space company, but it is enough to prove the software works. Their upcoming launch on a SpaceX rocket is the ultimate test. It is a bit ironic that the very company that said 'no' to the idea is now the one carrying the hardware into orbit. That is the beauty of the current launch market—SpaceX is so dominant as a bus driver that they have to carry their competitors' passengers.

What This Means for the Next Five Years

We are moving toward a future where 'the cloud' isn't just a metaphor for a server farm in Virginia. It will be a literal constellation of distributed nodes circling the planet. For builders, this means we need to start thinking about 'orbital-native' applications. If you had access to real-time, AI-processed global data with zero ground-station latency, what would you build?

It won't just be about better maps. It will be about automated supply chains that reroute ships based on real-time port congestion seen from space, or insurance models that trigger payouts the moment a flood is detected by an on-orbit model. The hardware is becoming a commodity; the intelligence on top of it is where the value will be captured.

The real opportunity isn't in building the satellite; it is in being the intelligence layer that makes the satellite useful in real-time.

My takeaway for founders is simple: watch the infrastructure. Whenever a new layer of the stack becomes 'open,' a massive wave of application-level innovation follows. If Satlyt succeeds in becoming the open standard for orbital AI, the barrier to entry for space-based services will collapse. You won't need to be a rocket scientist to build a space app; you'll just need to be a good developer.

Read the original at TechCrunch AI →