Audience
AI teams, ML engineers, researchers and startups that want GPUs without hyperscaler pricing or lock-in, especially teams that need to keep data in Europe.
About Compute with Hivenet
Hivenet Compute gives you on-demand RTX 5090 and RTX Pro 6000 GPUs, plus vCPU instances, billed per second while they run. Hivenet owns and operates the infrastructure, with data centres in France, the UAE and the US, so EU teams can keep workloads and data in Europe.
Use it for training and fine-tuning, inference, notebooks, rendering and batch jobs. Launch from ready-made templates in a few minutes, or automate everything through the public Compute API. If you'd rather not run the serving layer yourself, the Inference API gives you an OpenAI-compatible endpoint.
Teams get organisations with role-based access, prepaid credits, and S3-compatible object storage alongside hivenet compute.
Videos and Screen Captures
Dashboard
Product Details
Compute with Hivenet Frequently Asked Questions
Q: Where are Hivenet’s data centres?
Q: How does billing work?
Q: Are there egress fees?
Q: Which GPUs are available?
Q: How fast can I get a GPU running?
Q: Can I run LLM inference without managing servers?
Q: Is there storage?
Q: Does it work for teams?
Q: Is an RTX 5090 enough, or do I need an H100?
Q: What kinds of users and organization types does Compute with Hivenet work with?
Q: What languages does Compute with Hivenet support in their product?
Q: What kind of support options does Compute with Hivenet offer?
Q: What other applications or services does Compute with Hivenet integrate with?
Q: Does Compute with Hivenet have an API?
Q: What type of training does Compute with Hivenet provide?
Q: How much does Compute with Hivenet cost?
Compute with Hivenet Product Features
AI Cloud Providers
Hivenet is a European AI cloud that operates its own GPU infrastructure, with regions in France, the UAE and the US. It covers the full stack for AI teams: on-demand RTX 5090 GPUs from €0.75/hr, RTX Pro 6000 for enterprise workloads, vCPU instances, S3-compatible object storage and an OpenAI-compatible Inference API. Pricing is published and billed per second from prepaid credits, with no egress fees and no long-term commitment. Self-serve from the console in minutes, or automate everything with the public Compute API. Team organisations with role-based access make it workable for companies, not just individuals. A practical alternative to hyperscalers for teams that want predictable costs, and for European companies that want their AI workloads hosted in the EU.
AI Fine-Tuning
Fine-tune open models like Llama, Mistral, Qwen and Gemma on RTX 5090 GPUs from €0.75/hr, billed per second. With 32 GB of VRAM per card, LoRA and QLoRA fine-tuning of most 7B to 13B models fits on a single GPU, and multi-GPU instances (up to 8) cover bigger runs. Per-second billing means short experiments and hyperparameter sweeps cost exactly what they run for. Start from PyTorch or Jupyter templates with Hugging Face tooling, save your environment as an image, and relaunch it for the next run. Keep datasets and checkpoints in S3-compatible object storage with no egress fees. When the model is ready, serve it on the same platform with vLLM or a dedicated OpenAI-compatible Inference API endpoint.
AI Inference
Two ways to run inference on Hivenet. Self-managed: launch RTX 5090 instances from €0.75/hr with the vLLM template and serve your own models, billed per second. Managed: the Inference API gives you a dedicated, OpenAI-compatible endpoint, so existing code switches over by changing the base URL and you skip running the serving layer. The RTX 5090's 32 GB of GDDR7 and 1.79 TB/s of memory bandwidth make it a strong price-to-performance fit for serving 7B to 14B models, or larger quantised ones, and RTX Pro 6000 covers bigger models and sustained high-utilisation serving. No egress fees, so you aren't charged for the responses you send back to users. Regions in France, the UAE and the US for low latency and EU data residency.
AI Infrastructure
Hivenet Compute is GPU and CPU infrastructure operated by Hivenet end-to-end, with regions in France, the UAE and the US. Teams get on-demand RTX 5090 GPUs (32 GB GDDR7) from €0.75/hr, RTX Pro 6000 for heavier production workloads, and vCPU instances from 2 to 32 vCPUs. Everything is billed per second from prepaid credits, with no long-term commitment and no egress fees. Around the instances sit the pieces most AI teams need: S3-compatible object storage, an OpenAI-compatible Inference API for dedicated model endpoints, a public Compute API with an OpenAPI spec, and team organisations with role-based access. Workloads that need to stay in Europe can run in the French region. A good fit for startups, research labs and product teams that want production infrastructure without hyperscaler pricing, quotas or lock-in.
AI/ML Model Training
Train and experiment on RTX 5090 GPUs (32 GB GDDR7, 5th-gen Tensor Cores) from €0.75/hr, scaling to 8 GPUs per instance, with per-second billing so idle time doesn't land on your bill. Suited to computer vision, small and mid-sized language models, diffusion models and research prototypes, where a cluster of H100s is overkill. RTX Pro 6000 is available for larger memory needs. Templates for PyTorch, Jupyter and Docker get you training in minutes; save your environment and relaunch it identically. Keep datasets and checkpoints in S3-compatible object storage with no egress fees for moving them. Automate runs through the public Compute API. Research partners and customers include INRIA, Proteineer and the University of Arizona. Regions in France, the UAE and the US.
Cloud GPU
Rent RTX 5090 GPUs from €0.75/hr, billed per second, and stop paying the moment you stop the instance. Each RTX 5090 has 32 GB of GDDR7 memory and 1.79 TB/s of bandwidth, enough for most inference, fine-tuning and image generation work without paying for data-centre cards you don't need. Scale from 1 to 8 GPUs per instance, or move up to RTX Pro 6000 for larger models and sustained production load. Launch from templates for PyTorch, Jupyter, vLLM, ComfyUI and Docker, save your own images, and connect over SSH. No egress fees and no long contracts. Regions in France, the UAE and the US, so you can keep data in the EU when you need to.
Compute with Hivenet Verified User Reviews
Write a Review-
Probability You Would Recommend? 10 out of 10
"Good source of RTX 4090 GPUs for AI/ML needs" 5 out of 5 stars Posted 2024-10-16
Pros: - Super quick setup
- No BS/straightforward pricing
- Fast cold starts
- Reliable machines
- Good internet connectivityCons: I thought the out of the box image was a bit basic - just Ubuntu 24.04 w/CUDA. It was fine as I could quickly download and configure what I needed, but I would have appreciate some pre-packaged images.
The team replied and let me know they are quickly going to add some more.Overall: It was really simple to get started and use HiveCompute for my ML/AI needs. I had a side project using PyTorch and some transformers from HuggingFace. It was very quick to get everything setup with hiveCompute.
Read More... -
Probability You Would Recommend? 10 out of 10
"Powerful and reliable Cost-Effective AI Solution" 5 out of 5 stars Posted 2024-10-14
Pros: Is reliable GPU access for AI workloads at a fraction of the cost of traditional cloud providers and comes with a free trial. The ease of use is fantastic, with a good interface that lets us get started quickly. The support team is responsive and helpful. GPU's are powerful too.
Cons: Honestly, nothing to dislike. It meets our needs perfectly and saves us money. Although it is still being updated and such.
Overall: One of the best solutions for getting our ai modelling and workloads done. Can easily say our team will be using this over AWS and other expensive options.
Read More...
- Previous
- You're on page 1
- Next