Amazon Web Services is pouring record sums into artificial intelligence infrastructure on Earth, yet its top executive is in no rush to chase data centers in orbit. While rivals talk up space-based computing, AWS chief Matt Garman is signaling that the economics and engineering still do not add up, keeping orbital facilities in the realm of long-term speculation rather than near-term strategy.
His stance matters because AWS sits at the center of the cloud market and is shaping how governments and enterprises will run AI workloads for years. By casting orbital data centers as a distant prospect, Garman is drawing a sharp line between headline-grabbing visions and the practical constraints that still govern rockets, satellites, and the balance sheets that must support them.
Garman’s cool view on orbital data centers
Amazon Web Services CEO Matt Garman used a recent appearance at the Cisco AI Summit in San Francisco to tamp down expectations that server farms in orbit are around the corner. He described orbital data centers as “pretty far” from reality and stressed that, with current launch technology, the idea is “just not economical,” a judgment that cuts against the more optimistic timelines floated by space entrepreneurs and several startups that are already pitching in space compute services. In his view, the cost of sending racks of servers, storage systems, and cooling hardware into orbit, then maintaining them in a harsh environment, dwarfs any theoretical gains in latency or energy efficiency that advocates like to highlight, which is why he framed the concept as a long term curiosity rather than a serious near term investment for AWS in front of the Cisco audience, according to Amazon Web Services.
Garman’s skepticism is not limited to a single sound bite; it reflects a broader assessment of the entire stack that would be required to make orbital computing viable. He has pointed to the need for reliable launch capacity, hardened hardware, autonomous operations, and robust networking back to Earth, arguing that each layer still carries unresolved risk and cost. That is why he has characterized space data centers as “not a near prospect” and suggested that AWS will focus instead on scaling terrestrial regions and edge locations, a position that aligns with analysis that describes space facilities as dependent on an “entire stack getting built out” before they can compete with ground-based clouds, as highlighted in reporting on Why AWS.
Cost, launch limits, and the physics problem
Behind Garman’s caution is a blunt cost calculus that starts with rockets. Even as launch prices fall, he has argued that the expense of lifting dense, power-hungry data center gear into orbit remains prohibitive, especially when compared with building another hyperscale campus in a region with cheap land and renewable energy. Reports on his remarks note that he sees the economics of launching and maintaining such infrastructure as the central blocker, with the current generation of rockets and satellite buses unable to deliver the kind of mass, reliability, and serviceability that cloud customers expect from their critical workloads, a concern echoed in coverage that describes how the AWS chief warns about the cost and technical hurdles of orbital facilities.
The launch market itself illustrates the tension between ambition and constraint. SpaceX, which is central to many orbital data center pitches, achieved a record 165 Falcon 9 and Heavy launches in 2025, short of its own goal of 180, a gap that underscores how even the most aggressive launch provider is still capacity-constrained relative to its aspirations. Those 165 flights carried a mix of Starlink satellites, government payloads, and commercial missions, leaving limited room for the kind of heavy, specialized cargo that a serious orbital data center would require, which helps explain why Garman views talk of “a million” space data centers as premature, as detailed in an analysis of Falcon launch activity.
Musk’s orbital AI push and the startup contrast
Garman’s realism stands in sharp contrast to the enthusiasm of Elon Musk and a cluster of younger companies that are trying to turn orbit into the next cloud frontier. Musk has been promoting the idea of space-based AI processing that would sit alongside Starlink-style constellations, arguing that low Earth orbit could host specialized compute nodes for inference and training, and that future launch vehicles could make such deployments routine. That vision has helped “heat” the concept of space data centers in financial markets and on social media, even as AWS and other incumbents keep their capital focused on terrestrial build-outs, a divergence captured in coverage of how Musk heats up the orbital data center narrative.
Startups are also pitching orbital compute as a way to serve defense, Earth observation, and latency-sensitive applications, and some have announced plans for small demonstration payloads that would host AI accelerators in microgravity. Yet Garman has effectively poured cold water on these efforts, saying that despite multiple companies working on the idea, the technical and economic barriers remain “extremely challenging” and that AWS will not chase orbital hype at the expense of its core roadmap. Reports on his remarks emphasize that he sees these ventures as experiments rather than competitive threats, a view reinforced by analysis that notes how the number of startups is exploring space data centers while AWS keeps its distance.
Earthbound AI bets and Cisco’s distributed vision
Garman’s reluctance to chase orbital data centers does not signal any hesitation about AI itself; if anything, it highlights how aggressively AWS is investing on the ground. Amazon has announced an investment of up to $50 billion to expand AI and supercomputing capabilities for Amazon Web Services, a commitment that includes new data center regions, advanced networking, and custom silicon for training and inference. That figure, spelled out as “$50 billion” in company materials, underscores the scale of the terrestrial opportunity that AWS sees in serving US federal agencies and large enterprises, and helps explain why its leadership is wary of diverting capital into speculative orbital projects, as detailed in the company’s AI investment plans and the related breakdown of how Amazon Web Services will deploy that $50 billion across advanced compute and networking technologies.