Frequently Asked Questions
Do I need a static IP to host on Vast.ai?
A static public IP is strongly recommended but not strictly required. Vast.ai supports hosts behind NAT using their outbound-connection model, but hosts on dynamic residential IPs generally see lower utilization because the platform occasionally loses and re-establishes connectivity, which hurts reliability scores. If you have the option, a business-class connection with a static IP is worth the premium.
Can I host on a laptop GPU?
Technically yes for mobile 3080-class or 4090-class parts, but in practice it is not a sustainable plan. Laptop GPUs have aggressive thermal throttling to protect the chassis, and sustained GPU rental workloads run hot for hours at a time. The cooling system is not designed for that duty cycle, which leads to downclocking under load, poor benchmark scores relative to your listed GPU, and long-term thermal stress on the laptop. A desktop rig with proper airflow is a much better host.
Is this legal where I live?
For most people, renting out compute time from a home rig is legal but counts as self-employment or small-business income, which means you owe tax on the earnings. Some countries require a business registration above certain income thresholds. Some homeowner's insurance policies have exclusions for commercial activity on the premises. And some residential leases prohibit running servers. Talk to an accountant about the tax side, and check your ISP and lease terms before you commit. This article is not legal or tax advice.
Do ISPs care about hosting traffic?
Some residential ISPs prohibit running servers or commercial services in their terms of service. Enforcement varies widely — many hosts run for years with no issue — but if your ISP detects sustained 24/7 outbound connections or bandwidth use at the top of your plan, you may get a warning or a throttle. Read your ISP's acceptable-use policy. Business-class connections are generally server-friendly and often cheaper per Mbps at the same speeds once you factor in extras.
Do I need a dedicated machine, or can I host on my gaming PC part-time?
Part-time hosting works mechanically — you can stop the host agent when you want to game — but part-time hosts see substantially lower utilization. Vast.ai's matching algorithm prefers machines with high uptime because renters want predictability. A rig that's only online 12 hours a day can still earn, but the math usually says either commit to 24/7 hosting on a dedicated machine or skip it. If you want to test the waters with a gaming rig before buying dedicated hardware, that's a reasonable way to validate the model.
What about electricity costs?
Electricity is the biggest variable cost and the most location-dependent. The rough formula: (watts_under_load / 1000) × hours_rented × kWh_rate. Use a wall-plug meter to measure actual draw (GPUs often pull more than their stated TDP), and check your utility bill for your marginal kWh rate, not the average. In high-cost regions, electricity can eat the majority of revenue. See the profitability article for a full worksheet.
How much can I earn?
It depends heavily on your GPU tier, local electricity rate, marketplace conditions, and how well your rig maintains reliability. We deliberately don't quote a number here because current rates shift with supply and demand. Check the Vast.ai marketplace for live pricing on your GPU class, then work backward using the profitability template. Treat any income as variable and planning-worthy, not passive.
What's the difference between Vast.ai and RunPod?
Both are GPU compute marketplaces that run Docker containers on third-party hardware, but the host-side experience is different. Vast.ai is an open marketplace: anyone can sign up as a host and list a rig, with rates set by the host and filtered by renter search. RunPod operates more as a curated provider network — host onboarding is an application process, and payout structures differ. Vast.ai is easier to start with; RunPod may offer more stable utilization for the hosts they accept. Check each platform's current terms before choosing.
What hardware will I need besides the GPU?
At minimum: a modern motherboard with enough PCIe lanes, a CPU with at least 4 cores (8+ preferred), 16 GB+ of RAM, an NVMe SSD with 500 GB or more free, and a reliable PSU sized to 1.2–1.5x your peak draw. See the GPU selection article for detail on multi-GPU rigs, and the score rubric for the full baseline.
How long does it take to set up a new host?
If you already have Ubuntu 22.04 installed and a stable internet connection, you can go from OS install to listed on Vast.ai in roughly 2–4 hours. The bulk of that is driver and Docker setup — see the Docker + NVIDIA guide. Troubleshooting driver issues or networking quirks can extend that considerably; budget a full afternoon for your first machine and expect subsequent ones to go faster.
What happens if a renter tries to do something harmful on my rig?
Vast.ai workloads run in Docker containers, which isolates them from the host OS in the typical case. That said, containers are not a hard security boundary — kernel exploits and GPU driver vulnerabilities exist. Run your host on a separate network VLAN from your personal devices if possible, keep the host OS patched, and don't store anything sensitive on the same machine. Vast.ai also maintains its own abuse-reporting and renter-verification processes.
Can I host Windows workloads?
No. Vast.ai and RunPod both require Linux on the host (Ubuntu is the preferred distribution) because GPU passthrough to Docker containers depends on the NVIDIA Container Toolkit, which is a Linux-specific piece of infrastructure. If you're on Windows today, dual-booting Ubuntu 22.04 or 24.04 is the path forward. The setup guide walks through the Linux stack.
Check your rig in 60 seconds
Free, no sign-up, no data collected. Works in your browser or from the command line.
Run the Compatibility Checker →