Imagine launching data centers into orbit to power the future of Artificial Intelligence. This audacious vision isn’t just science fiction anymore, thanks to companies like Orbital, a new venture founded by Euwyn Poon, the entrepreneur behind the successful e-scooter company Spin. Poon, who sold Spin to Ford, has now secured a hefty $5 million seed round to build the foundational infrastructure for AI compute in space.
Orbital officially emerged from a16z’s prestigious Speedrun startup accelerator program in May, signaling a significant vote of confidence from a top-tier venture firm. Their ambitious goal is to perform AI inference in space, a critical step towards distributed, high-performance computing beyond Earth. This initial funding round also saw participation from a diverse group of investors, including Basis Set, Human Element, Wayfinder, Antler, Anti Fund, Ascent, Rubik, Zero Knowledge Ventures, LYVC, Feld Ventures, New Legacy, FNDR, UpHonest, and Asterisk.
According to Andrew Chen, a partner at a16z, Poon’s track record of scaling Spin—deploying 250,000 scooters across 100 cities—was a key factor in their enthusiastic support. Chen noted that Poon explored several ideas before settling on space data centers, demonstrating a methodical approach to identifying a impactful new frontier. His prior experience in rapid scaling is seen as invaluable for the complex challenges of aerospace development.
The Lure of Off-World Compute
The core proposition for space data centers is compelling: there’s an insatiable demand for AI compute, but deploying it on Earth faces limitations. These include significant environmental concerns, real estate constraints, and the sheer slowness of terrestrial infrastructure build-out. Moving compute to space promises benefits like limitless solar power and fewer regulatory hurdles, making it an attractive, albeit challenging, proposition.
However, the primary hurdle has always been the brutal economics of launching payloads into orbit. Current launch costs render the business case for large-scale space data centers unfeasible. This is where Orbital, like many of its nascent competitors, places a monumental bet on the evolving landscape of space travel.
Betting on Starship’s Promise
Orbital’s entire strategy hinges on the successful and regular operation of SpaceX’s Starship rocket. Founder Euwyn Poon explicitly stated, “We will get to full scale when Starship comes online.” He clarified that the cost of using current-generation rockets like the Falcon 9 simply “makes this not economically feasible” for their long-term vision.
For now, Poon and his Los Angeles-based team, comprising about a dozen experts with experience from Amazon LEO, SpaceX, and Northrop Grumman, are focused on immediate milestones. Their first step is a demo flight where they plan to fly an Nvidia Blackwell chip on a partner’s satellite. This critical test will evaluate Orbital’s innovative radiation shielding and thermal management technologies, essential for operating advanced hardware in the harsh space environment.
Looking further ahead, Orbital aims to launch its first dedicated data-processing spacecraft in 2028, equipped with Nvidia’s Space-1 Vera Rubin-class GPUs. At this stage, the company plans to generate revenue by performing “piece-wise inference work” with each launched satellite. This incremental approach allows them to build momentum and prove their capabilities while awaiting the full operational capacity of Starship.
A Vision for a Galactic Network
Ultimately, Orbital’s ambitious goal is to deploy a constellation of 10,000 satellites, collectively providing a distributed gigawatt of computing power. Each individual satellite is envisioned to contribute 100 kW of power to this vast network. For context, Elon Musk has indicated SpaceX’s AI satellites might produce up to 150 kW, while rival Starcloud anticipates larger 200 kW-rated spacecraft.
While some competitors, like a16z-backed Cowboy Space Company, are taking the extreme step of building their own rockets, and Blue Origin is also eyeing the space data center market with its New Glenn vehicle, Poon remains confident. He believes the sheer breadth of demand for AI applications ensures ample room for multiple players to succeed. There are “so many lanes for companies in our space to pursue,” he told TechCrunch, highlighting diverse approaches to AI workloads, hardware designs, and overall concepts for space-based data centers.
From Scooters to Satellites
The willingness of venture firms to back such capital-intensive, long-term projects marks a significant shift in the investment landscape. As Andrew Chen observed, a decade ago, when mobile apps dominated the tech scene, a project like Orbital would have seemed “crazy.” However, the current “energy and excitement” in capital markets for transformative space endeavors, largely spurred by SpaceX’s impact, makes such ventures viable today.
Poon’s journey to this groundbreaking idea began unexpectedly after leaving Ford, when he acquired an Nvidia A100 chip and co-located it in a Santa Clara data center. That firsthand experience of serving open-weight models convinced him of the immense value in delivering compute power in the burgeoning AI era. Now, with a clear vision and significant backing, he’s ready to tackle the monumental task of putting not just one, but thousands of GPUs into space.
Source: TechCrunch – AI
