Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
The NVIDIA H100 described as “heading to orbit” has already flown: Starcloud says its Starcloud-1 satellite launched in November 2025 and later ran AI workloads in space. The mission is a meaningful test of data-center-class computing beyond Earth—not proof that orbital data centers are ready to replace cloud infrastructure.
What launched—and what happened next
Starcloud-1 is a small technology-demonstration satellite developed by Starcloud, the startup formerly known as Lumen Orbit. It carried an NVIDIA H100, a data-center GPU designed for AI workloads, on a SpaceX Falcon 9 rideshare flight in November 2025. Spaceflight Now covered the launch, while Starcloud says the spacecraft became the first satellite to carry an H100 into space. Launch coverage · Starcloud-1 mission account
This was one GPU aboard an experimental satellite, not a commercial data center in orbit. A public satellite catalog lists the spacecraft at about 60 kilograms; that is an approximate catalog figure, not a complete official specification. SatNOGS satellite entry
Recommended Free Tools
Starcloud reports that the satellite ran Gemma, Google’s open model family associated with Gemini, and trained Andrej Karpathy’s nanoGPT model in orbit. The company describes the mission as demonstrating both inference and training. Those are significant operational claims, but they do not mean Google’s complete Gemini service—or a ChatGPT-scale public cloud service—was running from space. Starcloud’s results · NVIDIA’s account
#1 Best Overall
- PLEASE NOTE: Exporting an NVIDIA RTX Pro 6000 GPU outside the US requires strict adherence to the U.S. Export Administration Regulations (EAR) and issuance of an export license from the Bureau of Industry and Security (BIS). Compliance and Know Your Customer (KYC) screening may be required as a condition of order acceptance. [NVIDIA Blackwell Streaming Multiprocessor] The new SM features increased processing throughput, and new neural shaders that integrate neural networks inside of programmable shaders | DLSS 4: Multi Frame Generation ensures ultra-smooth frame pacing for lifelike simulations.
- [Double-Flow-Through Design] The RTX PRO 6000 Blackwell features a double-flow-through cooling design, optimizing efficiency and airflow to sustain peak performance under 600W power loads. | [5th Gen Tensor Cores] Deliver up to 3X the performance of the previous generation and support for FP4 precision for faster AI model processing times with reduced memory usage, enabling local fine-tuning of LLMs and generative AI | [4th Gen Ray Tracing Cores] Double the ray-triangle intersection rate of the previous generation to create photoreal, physically accurate scenes and immersive 3D designs with RTX Mega Geometry, which enables up to 100X more ray-traced triangles.
- [PCIe Gen 5] Support for PCIe Gen 5 provides double the bandwidth of PCIe Gen 4, improving data-transfer speeds from CPU memory and unlocking faster performance for data-intensive tasks like AI, data science, and 3D modeling. | [GDDR7 Memory] With 96 GB of GPU memory and 1.8 TB ps bandwidth, it can tackle massive 3D and AI projects, fine-tune AI models locally, explore large-scale VR environments, and drive larger multi-app workflows.
- [DisplayPort 2.1] Achieve unparalleled visual clarity and performance, driving high resolution displays at up to 8K at 240 Hz and 16K at 60 Hz. Increased bandwidth enables seamless multi-monitor setups while HDR and higher color depth support ensures superior color accuracy for precision work, such as video editing, 3D design, and live broadcasting.
- [Universal MIG] Divide a single RTX PRO 6000 Blackwell into multiple isolated instances, each with dedicated resources, allowing for concurrent execution of multiple workloads, optimized GPU utilization, and secure isolation of different applications or users. [WARRANTY] 3 YR Manufacturer's Warranty. Bulk OEM Packaging. Retail Packaging is NOT included.
Why send an H100 into space?
An H100 offers substantially more AI compute capability than the conventional low-power processors commonly used for basic spacecraft control and data handling. It is not, by default, a radiation-hardened spacecraft computer; the point of this demonstration was to test data-center-class hardware and AI software in an orbital environment. NVIDIA describes the mission as an early test of computing beyond Earth and of processing data closer to where it is collected. NVIDIA H100 overview · NVIDIA on Starcloud
One proposed use is satellite imagery analysis. A spacecraft collecting Earth-observation data could run an AI model onboard to flag a wildfire, detect a change, or prioritize images for transmission. Sending a concise alert or selected image instead of every raw frame could ease a downlink bottleneck. This is most attractive when the data originates in space and the result is useful before the next practical opportunity to contact a ground station.
Onboard computing does not automatically mean faster, cheaper, or more reliable service. The result depends on the sensor’s output, model size, available power, thermal limits, orbital contact windows, and the capacity and scheduling of the satellite’s communication links. A compute-heavy result still has to reach a user on Earth.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Rank #2
- NVIDIA Ampere Architecture-based CUDA Cores - Double-speed processing for single-precision floating point (FP32) operations and improved power efficiency provide significant performance improvements for graphics and simulation workflows, such as complex 3D computer-aided design (CAD) and computer-aided engineering (CAE), on the desktop.
- Second-Generation RT Cores - With up to 2X the throughput over the previous generation and the ability to concurrently run ray tracing with either shading or denoising capabilities, second-generation RT Cores deliver massive speedups for workloads like photorealistic rendering of movie content, architectural design evaluations, and virtual prototyping of product designs. This technology also speeds up the rendering of ray-traced motion blur for faster results with greater visual accuracy.
- Third-Generation Tensor Cores - New Tensor Float 32 (TF32) precision provides up to 5X the training throughput over the previous generation to accelerate AI and data science model training without requiring any code changes. Hardware support for structural sparsity doubles the throughput for inferencing. Tensor Cores also bring AI to graphics with capabilities like DLSS, AI denoising, and enhanced editing for select applications.
- Third-Generation NVIDIA NVLink - Increased GPU-to-GPU interconnect bandwidth provides a single scalable memory to accelerate graphics and compute workloads and tackle larger datasets.
- 48 Gigabytes (GB) of GPU Memory - Ultra-fast GDDR6 memory, scalable up to 96 GB with NVLink, gives data scientists, engineers, and creative professionals the large memory necessary to work with massive datasets and workloads like data science and simulation.
Space solar power is abundant, not uninterrupted
Starcloud’s longer-term vision is to build solar-powered orbital data centers. In orbit, solar arrays can avoid terrestrial weather and may receive sunlight for much of an orbit. But “solar-powered” does not mean continuous power: eclipses can interrupt generation, and output depends on orbit, spacecraft orientation, array size, degradation, and power-system efficiency. Batteries and power electronics add mass and additional components that can fail.
The company has described eventual facilities reaching gigawatt scale, with solar arrays and radiators roughly four kilometres across. Those dimensions and ambitions are company proposals, not operating infrastructure or independently validated economic projections. Starcloud’s vision · NVIDIA’s description of the concept
Space is not an easy place to cool a GPU
A GPU turns electrical power into heat. On Earth, cooling systems move that heat into air or liquid and then reject it to the environment. In vacuum there is no air for fans to push over a heatsink; heat must ultimately be carried to radiators and emitted as infrared radiation. Space’s low ambient temperature does not make heat disappear.
Rank #3
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
Radiator area, orientation, exposure to sunlight and Earth’s infrared radiation, and the spacecraft’s ability to carry heat away from the GPU all matter. Higher compute duty cycles mean more heat to reject. Starcloud-1’s reported AI runs show that workloads were performed, but public mission material does not establish long-term thermal margins under a commercial facility’s sustained, dense operating load.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Why one successful demonstration does not settle the business case
- Radiation and reliability: Space radiation can cause memory errors, logic faults, latch-ups, or permanent damage. Error correction, watchdogs, redundancy, checkpointing, and recovery procedures help, but the public results do not provide a complete, independently audited account of the H100’s error rates or multi-year reliability.
- Launch and replacement: Hardware, arrays, radiators, batteries, and shielding all add mass that must be launched. A failed component is not normally something a technician can swap out as they would in a terrestrial data center. Launch vibration, acoustic loads, qualification, and replacement missions add cost and schedule risk.
- Communications: Onboard inference may shrink the data that needs to be downlinked; it cannot eliminate ground stations, command and data links, network scheduling, authentication, or latency. A satellite with substantial compute can still be constrained by when it can communicate.
- Upgrades and obsolescence: AI hardware and models change quickly. A satellite can remain in orbit after its processor becomes less competitive, while upgrading it is far harder than replacing servers on Earth.
- Operations and regulation: A larger fleet would have to manage orbital traffic, collision avoidance, end-of-life disposal, spectrum access, cybersecurity, and applicable national rules.
Training a small model in orbit is technically notable, but it does not establish that training is the best orbital workload. Inference on data generated by the satellite may have a clearer edge-computing case, since it can reduce unnecessary transmission. For either use, a real comparison needs power and duty-cycle figures, hardware lifetime, useful output per unit of launch mass, network costs, ground operations, and the cost per useful result.
Starcloud has argued that orbital facilities could reduce energy-related costs or emissions compared with terrestrial data centers. Such claims remain a company thesis: a fair lifecycle comparison would also have to account for spacecraft and launch manufacturing, replacement, ground infrastructure, and operations. The public information cited for Starcloud-1 does not provide a complete cost model, sustained utilization data, or a basis for concluding that orbital computing is commercially cheaper.
Rank #4
- Standard Memory: 40 GB
- Host Interface: PCI Express 4.0
- Cooler Type: Passive Cooler
- Product Type: Graphics Card
What comes next?
Y Combinator and a KPMG industry summary have described a second Starcloud satellite as planned for October 2026, with reporting that future hardware may include NVIDIA Blackwell and multiple H100s. These are reported plans, not confirmation that a launch has occurred or that those configurations are final. Starcloud’s Y Combinator profile · KPMG industry summary
The harder milestone is not merely putting more GPUs in orbit. It is demonstrating a reliable, maintainable, networked system whose useful output justifies the hardware, launch, power, heat-rejection, and operations costs. Starcloud-1 is a step toward testing those questions, not their answer.
Free tools Windows power users keep installed
One-click scans. No signup required.
Could you use Starcloud-1’s GPU?
No ordinary self-serve rental or consumer access path for Starcloud-1’s orbital H100 is identified in the cited material. Starcloud is developing orbital infrastructure; that is different from offering a public cloud GPU instance. Organizations seeking AI compute today should evaluate terrestrial services instead. For example, Google Cloud GPU infrastructure, Amazon EC2 accelerated instances, and Azure GPU virtual machines describe land-based options. These are not equivalent to a spacecraft GPU, and availability, pricing, and fit depend on provider, region, configuration, and workload.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
