Published: June 03, 2026, 15:56 UTC
Impulse Space Raises $500M Series D as Orbital Tugs Prep for a Future That May — or May Not — Include Space Data Centers
Impulse Space, the in-space mobility company founded by former SpaceX propulsion chief Tom Mueller, closed a $500 million Series D funding round on Tuesday, led by 137 Ventures and BANNER VC. The company has now raised over $1 billion in total — including a $300 million Series C just over a year ago — signaling sustained institutional conviction that orbital transportation infrastructure is becoming central to the space economy.
The Series D, announced June 2, will fund continued development of Impulse’s two core products: Mira, a maneuverable spacecraft platform for satellite deployment and payload hosting, and Helios, a high-energy kick stage designed to haul payloads from low Earth orbit to geostationary orbit. Helios is particularly relevant as GEO satellite communications operators face growing pressure to get assets on-station faster and improve capital efficiency.
“Demand is strong across civil, commercial, and national security domains,” said Eric Romo, Impulse Space’s president and COO, in the company’s announcement. The company’s Caravan GEO rideshare missions are slated to begin in 2027, offering fractional orbital transfer services that could reshape how satellite operators plan constellation deployment.
The national security angle is noteworthy. Impulse has been steadily growing its relationship with the U.S. Space Force, particularly around dynamic space operations and responsive launch — the ability to maneuver assets, reposition satellites, and rapidly respond to orbital threats or capability gaps. In an era of growing great-power competition in space, mobility is increasingly seen as a strategic imperative, not just a commercial convenience.
Beyond Earth orbit, Impulse is working on a lunar lander concept capable of delivering up to three tons of cargo to the Moon’s surface — a play for the growing cislunar logistics market that NASA’s Artemis program and its commercial partners are building.
“In-space mobility is becoming foundational infrastructure for the space economy,” said Adam Ramada of BANNER VC in a statement. “As activity in orbit increases, the ability to move payloads efficiently between orbits is no longer a nice-to-have — it’s a requirement.”
That raises a question: If Impulse’s orbital tugs become the freight trucks of cislunar space, what cargo are they hauling?
The Infrastructure Play — and the Data Center Dimension
The answer that’s getting the most attention in 2026 is data. Specifically, orbital data centers — server farms in space that would leverage abundant solar energy, bypass terrestrial grid constraints, and escape the growing wave of community opposition to land-based AI data center construction.
The market signals are unmistakable. According to Via Satellite, U.S. data center capital spending rose 70% from May 2023 to May 2024, and a Lawrence Berkeley National Laboratory study projects that data center energy consumption could double or triple by 2028, consuming up to 12% of U.S. electricity. A World Economic Forum report published June 2 notes that 70% of Americans oppose local AI data center construction. That combination — surging demand, limited grid capacity, public resistance — has pushed some of the biggest names in tech to look skyward.
SpaceX has reportedly filed for a constellation of up to one million satellites for orbital compute. Google is exploring TPU clusters in space. Starcloud, a venture-backed startup, has proposed an 88,000-satellite constellation for on-orbit computing. And in the most prominent signal yet, Google and SpaceX are in preliminary discussions about partnering on orbital data centers, with a focus on AI inference workloads, according to SignalFeed.
The pitch is seductive. Launch costs are falling, powered by SpaceX’s Starship reusability. Solar energy in space is continuous and roughly seven times more intense than on Earth’s surface. No land-use permits, no water rights battles, no zoning fights. For AI companies facing an insatiable appetite for compute and an increasingly hostile regulatory environment on Earth, the sales pitch writes itself: put the servers where the energy is, and let the vacuum do the rest.
The Physics Problem Nobody Wants to Talk About
There’s just one catch. The vacuum.
“The assumption that space is cold and therefore cooling is easy is dangerously wrong,” reads a World Economic Forum analysis of the problem. In a vacuum, there is no air to conduct heat away. No water to circulate through cooling towers. Waste heat can only escape via infrared radiation — literally glowing it away.
This is orders of magnitude less efficient than terrestrial cooling. A Starcloud white paper cited by the WEF calculates that a two-sided radiator operating at 20°C emits only 633 watts per square meter. By comparison, water cooling on Earth sheds heat over 1,000 times faster. A one-megawatt orbital data center — already a thousand times smaller than a typical hyperscale terrestrial facility that runs in the gigawatt range — would need 1,600 square meters of radiator surface area. That’s the size of a hockey rink. In orbit.
And those radiators can’t just be deployed anywhere. The thermal design problem is genuinely wicked: solar arrays want to face the Sun to generate power, while radiators want a clear view of cold space to dump heat. These are directly conflicting pointing requirements. The radiators themselves must be thick enough to conduct heat efficiently, meaning they’re heavy — and every kilogram launched to orbit costs money.
To put some numbers on it: NASA’s External Active Thermal Control System on the International Space Station rejects roughly 70 kilowatts of heat through radiator wings that mass about seven metric tons. Scaling that to a one-megawatt orbital data center — still a tenth the scale of a terrestrial facility — implies roughly 100 tons of radiators alone. Starship’s advertised payload capacity to LEO is around 100 tons, meaning a single data center module could consume an entire Starship launch just for its thermal management system. That’s before you account for the servers, the power systems, the structural framework, and the propulsion to maintain orbit.
The European Union’s Ascend study, led by Thales Alenia Space, proposed a notional 10-megawatt orbital data center consisting of 13 satellites arranged in a 200-meter by 83-meter formation. That’s technically impressive, but the study’s own findings are sobering: the concept would require rockets roughly ten times less carbon-intensive than today’s launchers to be environmentally friendlier than terrestrial alternatives. And scaling that to even 200 megawatts — a fraction of a single hyperscale campus — would require 200 such orbital formations and roughly 200 dedicated launches.
“Launching hardware remains extremely expensive,” Dr. Domenico Vicinanza of Anglia Ruskin University told the BBC. “Each kilogram costs thousands of dollars.”
The Verdict: Niche, Not Revolutionary
None of this means orbital data centers are impossible. Avi Shabtai, CEO of radiation-hardened computing specialist Ramon.Space, has called orbital compute “the opportunity of the century,” while candidly acknowledging the “hostile environment” of space — radiation, thermal extremes, energy constraints, and the sheer difficulty of transportation. NVIDIA is collaborating with a half-dozen space companies including Aetherflux, Axiom Space, Kepler, Planet, Sophia Space, and Starcloud on orbital compute architectures. There is real engineering momentum.
The most plausible near-term use cases are edge computing applications that genuinely need to be in space — on-orbit data processing for satellite imagery, real-time AI inference for autonomous spacecraft operations, and latency-sensitive workloads for military space assets. These don’t need megawatt-scale facilities; they need ruggedized, radiation-hardened servers that can process data where it’s collected, avoiding the bandwidth bottleneck of downlinking everything to Earth.
For the hyperscale AI training workloads that are driving data center demand today, however, the economic equation heavily favors terrestrial infrastructure — even with grid constraints, community opposition, and rising energy costs. A gigawatt-class data center in Arizona or Virginia is a construction project. A gigawatt-class data center in orbit is a generational engineering challenge, requiring hundreds of launches, tens of thousands of tons of hardware, and thermal management systems that don’t yet exist at scale.
Impulse Space is building genuinely important infrastructure. If there is a future in which orbital compute nodes matter — for defense, for edge AI, for low-latency space-based services — companies like Impulse that can move payloads efficiently between orbits will be essential. But their business model doesn’t depend on the orbital data center thesis being right at hyperscale. It depends on doing one thing well: getting stuff from one orbit to another. That’s a bet on mobility, not on megawatts.
The data center in space may come. But it will arrive much later, much smaller, and much more expensively than the hype suggests. For now, Impulse Space’s $500 million is a bet on the here and now of the space economy — not on a server farm among the stars.