01
Hostman
Cloud 6 · 8 × 16 GB · 320 GB NVMe · Europe
✓ Listed for “AI and ML”✓ Plan covers 8 cores and 16 GB RAM✕ Over budget by $12
Match by workload
VPS for AI and ML workloads
Model inference, n8n with AI, vector databases, your own LLM front-end
How many resources you need
Why that many
AI tasks on a virtual server usually mean not training models but inference, orchestration and vector search. Training needs a GPU and will not run on an ordinary VPS. But the scenarios “your own front-end to a language model”, “a vector database for document search” and “automation with calls to an external API” live well on CPU servers, and they need memory for the indexes
01 · Minimum
2 × 4 GB
disk from 40 GB
A layer to an external model API, a chat interface, simple agents
02 · Workingtypical choice
4 × 8 GB
disk from 80 GB
A vector database, document search, a task queue
03 · With headroom
8 × 16 GB
disk from 160 GB
Local inference of small models on the CPU, large indexes
What to check before paying
- Whether you need a GPU: local inference of large models on the CPU is not cost-effective
- Network speed and the traffic limit, if the model is called over an external API
- The ability to grow memory quickly for a growing vector index
- Hourly billing for experiments that live a few days
How this ties into the matching. The calculator below is already set up for this workload: 8 cores and 16 GB of memory from the recommendation are filled in. You can change the values and the match recalculates instantly. A workload fit adds +25 points to a provider, the full formula
Matching
Find a server for a ai and ml
Server matching
108 plans with a verified price, data checked August 17, 202601 · Workloadwe suggest 8 × 16 GB
Model inference, n8n with AI, vector databases, your own LLM front-end
02 · Location
03 · Memory16 GB
1 GB64 GB
04 · Cores8 cores
116
05 · Budgetup to $25 / month
$5no limit
06 · Requirements
Provider commission is not part of the formula. How the score works
Matching result20 matching plans
02
Cherry Servers
Cloud VPS 1 · 1 × 1 GB · 20 GB SSD · Europe
✓ Listed for “AI and ML”✕ Closest plan is below your request: 1 × 1 GB✓ Within budget
03
RUVDSRussia
Powerful 1 · Europe · 2 × 4 GB · 40 GB SSD · Europe
✓ Listed for “AI and ML”✕ Closest plan is below your request: 2 × 4 GB✓ Within budget
04
DataPacket
VServer 1 GB · 1 × 1 GB · 50 GB NVMe · USA
✓ Listed for “AI and ML”✕ Closest plan is below your request: 1 × 1 GB✓ Within budget
05
VPSDime
Linux VPS 6 GB · 4 × 6 GB · 30 GB NVMe · USA
✕ “AI and ML” is not listed by the provider✕ Closest plan is below your request: 4 × 6 GB✓ Within budget
06
RackNerd
KVM 1 GB · 1 × 1 GB · 20 GB SSD · USA
✕ “AI and ML” is not listed by the provider✕ Closest plan is below your request: 1 × 1 GB✓ Within budget
Some links are affiliate links: if you sign up we earn a commission at no extra cost to you. It does not affect the ranking, the matching formula is open
Plans
Suitable plans from the base
Selected are the ones that cover the recommended configuration: at least 8 cores and 16 GB of memory
| Provider | Plan | CPU | RAM | Disk | Price per GB | Verified | Price |
|---|---|---|---|---|---|---|---|
| Hostman | Cloud 6 | 8 | 16 GB | 320 GB NVMe | $2 | August 3, 2026 | $37 |
Providers
Who claims this workload
HO
A managed cloud that deploys apps from your repository in a few clicks. A fit for teams without dedicated DevOps: updates, monitor…
affiliate linkhourly
Plans from$4
Focused on performance and dedicated servers for production. A choice for heavy workloads where stable capacity matters more than…
affiliate link
Plans from$5
RU
A provider with many data centers and dedicated products for 1C, trading and GPU. Notable in the corporate segment.
affiliate linkhourly
Plans from$14
Cloud VPS with hourly billing and broad geography: Europe, the US and Asia. Handy when you need a specific host country and flexib…
affiliate linkhourly
Plans from$4
Questions
Common questions about a ai and ml
Can a language model run on an ordinary VPS
Small quantized models run on the CPU and work, but slowly: a few tokens per second. Interactive use needs a GPU. A practical setup at the start is to call an external API and keep the logic, history and vector search on your own server
How much memory does a vector database need
It depends on the dimension and the number of vectors. A guideline for estimation: a million vectors of dimension 768 in a four-byte representation take about three gigabytes for the vectors alone, without the index and metadata. Plan memory with a twofold headroom