BW Businessworld last spoke with [color=hsl(0,75%,60%)]Awais Ahmed, Founder and CEO at Pixxel, in January 2025, when the company launched India’s first private satellite constellation aboard a SpaceX rocket from the Vandenberg Space Force Base. Much has changed since. In early May, Pixxel announced a partnership with Sarvam to build India’s first orbital data centre satellite and, in the same week, secured a contract with the US National Reconnaissance Office (NRO). We caught up with Ahmed to understand how Pixxel’s satellite business, global contracts and orbital data centre plans are converging into its next chapter. Excerpts:
When we last spoke, ‘Pathfinder’ was known primarily as a hyperspectral imaging satellite. When did the orbital data centre idea come in?
It’s not just another hyperspectral satellite. If it were just a hyperspectral satellite, it would not have been as powerful or as heavy. The idea is that, in general, we are moving towards bigger, larger, and more powerful satellite platforms.
There are two business lines that Pixxel focuses on. One is our own satellite constellation and our ‘Aurora’ platform for analysis, that’s the planetary intelligence piece, where we own the sensors, the satellites, the data, and we analyse it through Aurora for a health monitor of our planet. The other is where we design, build, and provide satellites to organisations that need their own satellites or constellations — the Air Force, Cosmoserve, and others. The backend and infrastructure is the same for both. So the idea was: can we hit two birds with one stone? We want to show that we have more powerful, more capable platforms, and at the same time extend our planetary intelligence vision.
"We want to show that we have more powerful, more capable platforms, and at the same time extend our planetary intelligence vision"
The (Pathfinder) satellite has a hyperspectral camera as an extension to our own constellation. But in addition, now, there will be powerful data centre GPUs on the satellite itself. These will be ground data centre-grade GPUs, not the low-power edge processors that are generally launched. That’s the real change.
What makes putting ground data centre-grade GPUs in space so difficult?
They take a lot more power, and they disseminate a lot more heat. So we have to figure out how to provide power continuously, and how to radiate that heat out into the vacuum of space. It’s a much larger and more capable satellite platform because of that. The GPU cluster itself, including thermal management, will be about 30 to 40 kilograms of the satellite’s 200 kilograms.
And what can these orbital data centres actually do in practical terms that can’t be done on earth?
Use case one is analysing data generated in orbit itself. The camera on a satellite captures what’s happening on the planet, and instead of running only the few things edge computing devices can handle, we now run it through an entire geospatial compute foundation model that can do a whole lot more. So, the capability goes up significantly. But only about 10 to 15 per cent of the data centre’s utility is taken care of by that imaging use case. More than 75 to 80 per cent can be for training very heavy AI models and showing it can be done in orbit.
However, the challenge is that when you’re training something, the GPUs have to be on for a long while. And not just 30 seconds, not a minute. You have to make sure it’s working across the entire orbit, that there’s enough power, and that there are enough backups in place so that if a bit flip or a radiation event happens, we can go back to where training was and continue without losing a beat.
What kind of GPUs are we talking about?
We are in touch with Nvidia and a few others. These would be ground data centre-grade. Think an H200 or B100. There’s also Nvidia’s Vera Rubin space variant, which is built specifically for space applications, but it’s not quite ready for procurement yet. In the next few months, it should be. So as of now, the H200s, H100s, B100-grade GPUs are what we’re building around. They’re fairly swappable and the power and other things change, but we’re building it for that modularity. For the first pathfinder, there will be at least two fully capable GPUs.
Beyond Sarvam, are you in conversation with others to use this platform?
We are, but I can’t name them until something is closer to announcement. What I can say is there are two aspects to this. One is proving the technology for global use. The other is: how can India not be left behind? How can we have a sovereign stack? An Indian-built satellite launching from India, with an Indian-built and trained model on the satellite — so that if, as a government or as our armed forces, there is a need for data centres in orbit, we are capable of doing it.
On launch vehicle, will it be a SpaceX rocket again?
We keep all our options open until about three to four months before launch. Generally, it's either SpaceX or Isro, those are the two major options. We go with whichever has a launch available and on time.
Let’s talk about Pixxel’s commercial journey. The Firefly constellation is operational now. The NRO contract was just announced. How competitive are those processes in the US?
It all comes down to how good your product is. We are the only provider of five-metre hyperspectral data in the world. When we launched the Firefly constellation last year, it was the world’s first and highest-resolution hyperspectral satellite constellation — not something we took from someone else and adapted, but genuinely the world's first. That’s why when NASA looks at it, or the NRO looks at it, it becomes easier for us as a non-US-native company to work with them. The US defence establishment works with allied countries such as the UK, European companies, India too. In the end, the process is very straightforward. You apply through the right channels, it goes through evaluation, and you find out if you got it. There’s nothing special we did except have access to a dataset no one else can provide.
How is the commercial side of your business shaping up in terms of geographies and industries?
Geographically, most of our work is India and the US, fairly similar in volume. The government tends to do heavy lifting at the early stage in both countries. But from a commercial standpoint, the US has far more companies with the paying capacity compared to India. Europe is not far behind the US commercially, they’re fairly close. Asia, South America and Africa are much smaller percentages.
In terms of industries, agriculture is the single largest, closely followed by mining and oil & gas. Under agriculture, I include agri-finance as well. And then there’s government: forestry and environment on one side, and defence and intelligence on the other, where we provide data on what's happening in geopolitically tense zones.
Are you profitable at the constellation level?
This year we will be. From a gross margin standpoint, the data business runs at 70 to 90 per cent depending on scale. The largest cost was the capex for launching the satellites, we’ve already done that for six satellites. For the next seven years, the lifetime of these satellites, we are generating revenue every month at very high margins. Ground stations, cloud storage, the people keeping it running — these are a minuscule fraction of the contracts we’ll have.
“We should be EBITDA positive in 2026 and cash flow positive by late 2027 or early 2028 at the latest”
Last year, we were still spending a lot of capex and the launches had to happen. This year, it’s about revenue generation with a few operational expenditures. We should be EBITDA positive in 2026 and cash flow positive by late 2027 or early 2028 at the latest. We build our satellites for a seven-year lifetime; most satellites are built for three. That makes a meaningful difference to the economics.
Is your Honeybee constellation on track for this year?
The first of the launches should happen this year, and we’ll follow it up with many more next year.
And the EO satellite consortium contract — you led that with Piersight, Satsure, Dhruva Space. Where does that stand?
Work started in January 2026, when we signed the contract. Launches are expected from 2028, maybe late 2027. The roles are clear: Pixxel handles the ultra-high resolution sub-metre satellites and the hyperspectral satellites. These will be seven of the total twelve. PierSight handles synthetic aperture radar. Satsure handles multi-spectral. Dhruva Space handles a lot of the components and the ground station infrastructure. None of us step over each other. As long as we each deliver by 2028, we’ll have it up and running.
You have Honeybee constellation, the EO consortium and Pathfinder in works. All projects are running in parallel. Are there capacity bottlenecks at your facility?
Right now we have a 20 to 25 satellite in-parallel capacity at the Megapixel facility, and that’s already fully taken. So the Gigapixel facility, which will be ready before the end of this year, is being built to take us to 100 satellites in parallel. In the most efficient cycle, that’s 300 to 400 satellites a year.
And as you expand, how many people are you hiring?
We’re at 260 globally right now and plan to add 50 to 75 more this year.
Finally — where does the Indian private space sector stand in 2026?
All the right ingredients are there. The government is being a customer to startups, providing policy frameworks, IN-SPACe as a single-window clearance, Isro facilities for testing. Every year many new startups get incorporated. Skyroot is targeting an orbital launch very soon. That will be a big boost to the ecosystem. We launched our constellation last year and are expanding it this year and next.
"We are still three to five years away from being close to China’s capabilities, and further from the US"
But let’s keep things in perspective. We are still three to five years away from being close to China’s capabilities, and further from the US. The companies that will define this decade are the ones moving from R&D to commercialisation. Either you focus on India as the market, which is a valid approach and there’s enough to do here — or you build something so unique you can sell internationally at scale. Both paths are valid. Both will take off. What I can say is the initial signs of success are there. |
