# CloudGPU > cloudgpu.app | Compute; 1 record: free tier, limits, identity cost and evidence. HTML https://fshy.ajam.dev/p/cloudgpu/ | JSON https://fshy.ajam.dev/api/v1/entity/cloudgpu.json | how to read a record: https://fshy.ajam.dev/llms-full.txt ## CloudGPU - `compute/cloudgpu` | paid-only | evidence documented | binding volume-quota | category GPU cloud - needs card no, phone no, ID no, institution no | verified 2026-11-03T07:36:21.177Z (live) | stale on 2026-12-04 - url https://cloudgpu.app/pricing - free tier: no free tier and no signup credit: a prepaid balance from a USD 5 minimum top-up, spent per minute on GPU rentals at list rates, with the A100 80G at USD 1.59/hr and the RTX 4090 24G at USD 0.39/hr. From 3 November 2026 a single top-up of USD 100 or more earns a USD 10 bonus, once per account. | unit: none | metered on: wall-clock | read on: 2026-11-03 | source: [1] | note: none - per-object cap: not applicable | source: [1] | note: the ceiling that matters is the card chosen on the deploy page - retention: Destroying an instance deletes everything stored on it, and an instance is destroyed if the balance runs out. | source: [2] | note: a machine record: what is lost is the instance and its local storage; copy results off before the instance ends - egress: no egress charges: the vendor states 'No hidden fees. No egress charges. Stop anytime.' | metered: no | free allowance: unmetered: no egress charge is stated | source: [2] - identity: email | detail: Sign-up is an email address or a Google account and no phone number is needed. There is no signup credit, so use needs a prepaid top-up of at least USD 5 by card, by USDT on TRC-20, or by bank transfer for business customers. - interface: ssh, rest - limits: - volume-quota (binding): prepaid balance only: no signup credit, a USD 5 minimum top-up, and instances are destroyed when the balance reaches zero [1] - limits note: the binding count is the prepaid balance; nothing is free, so the cost of one run is the list rate multiplied by the minutes billed - binding note: the binding count is the prepaid balance: there is no free allowance, the minimum top-up is USD 5 and a balance of zero destroys the instance - interface detail: root SSH and JupyterLab on the instance templates, and an OpenAI-compatible API at https://cloudgpu.app/v1 for the per-token models; read from the vendor pages, not probed - tooling: root SSH and JupyterLab on every instance; one-click templates from the console; an MCP server that lets an MCP client quote a price, rent a GPU after approval, run a job and shut the GPU down - composes with: none: no free lane to chain. The smallest purchase is a USD 5 top-up, and results must be copied off the instance before it is destroyed. - reliability: the vendor page states a public beta with no uptime guarantee and no service credits; capacity is sourced from third-party suppliers and the supplier behind an instance can change - notes: The vendor price table read on 2026-11-03 names 20 cards and prices the A100 80G at USD 1.59/hr and the RTX 4090 24G at USD 0.39/hr; it names no H100, L40S or L4. The 4090 also comes as a 48 GB card at USD 0.69/hr. The aggregator computeprices.com lists the same A100 SXM and RTX 4090 rates for this host under the region Asia-Pacific. - evidence: - [1] 2026-11-03: the A100 80G list rate is USD 1.59/hr on the vendor pricing page read on 2026-11-03 | capture: sha256:d912a60f2d87029f7ce65348f62d9be471c821fb95476d0458a331bdc4f50e68 | quote: | A100 80G | $1.59/hr | - [1] 2026-11-03: the RTX 4090 24G list rate is USD 0.39/hr on the vendor pricing page | capture: sha256:d912a60f2d87029f7ce65348f62d9be471c821fb95476d0458a331bdc4f50e68 | quote: | RTX 4090 24G | $0.39/hr | - [1] 2026-11-03: the nearest H100 row is the H800 at USD 1.79/hr, an H100 variant with NVLink capped at 400 GB/s; the table names no H100, L40S or L4 | capture: sha256:d912a60f2d87029f7ce65348f62d9be471c821fb95476d0458a331bdc4f50e68 | quote: | H800H100 variant, NVLink capped at 400 GB/s | $1.79/hr | - [1] 2026-11-03: the service is a public beta with no uptime guarantee | capture: sha256:d912a60f2d87029f7ce65348f62d9be471c821fb95476d0458a331bdc4f50e68 | quote: CloudGPU is in public beta. There is no SLA: no uptime guarantee and no service credits. - [2] 2026-11-03: no egress charges are stated | capture: sha256:f3cd389bb9503e0b570d1231f02dd02c34f373c0fe2f432480b063218e709219 | quote: * No hidden fees. No egress charges. Stop anytime. - [2] 2026-11-03: GPU time is billed per minute with a 1-minute minimum | capture: sha256:f3cd389bb9503e0b570d1231f02dd02c34f373c0fe2f432480b063218e709219 | quote: Per minute, rounded up, with a 1-minute minimum. - [2] 2026-11-03: instance storage is lost on destroy and a zero balance destroys the instance | capture: sha256:f3cd389bb9503e0b570d1231f02dd02c34f373c0fe2f432480b063218e709219 | quote: Destroying an instance deletes everything stored on it, and an instance is destroyed if the balance runs out. - [2] 2026-11-03: instances come with root SSH and JupyterLab | capture: sha256:f3cd389bb9503e0b570d1231f02dd02c34f373c0fe2f432480b063218e709219 | quote: Root SSH and JupyterLab - [3] 2026-11-03: no phone number is needed; sign-up is an email address or a Google account, and payment is by card, USDT or bank transfer | capture: sha256:bdbd7764dfc09b8b842ec4684fb6c91cfdbdc49350d573840fa27c1be5d853a8 | quote: No. Sign-up is an email address or a Google account, and payment is a card, USDT or a bank transfer. - [3] 2026-11-03: the minimum top-up is USD 5 | capture: sha256:bdbd7764dfc09b8b842ec4684fb6c91cfdbdc49350d573840fa27c1be5d853a8 | quote: The minimum top-up is $5. - [3] 2026-11-03: USDT on TRC-20 is accepted from USD 5 | capture: sha256:bdbd7764dfc09b8b842ec4684fb6c91cfdbdc49350d573840fa27c1be5d853a8 | quote: USDT on the TRON network (TRC-20) is accepted, minimum $5 - [4] 2026-11-03: there is no signup credit | capture: sha256:78cd8c2ec45d24746a206dc7cf4c5a1881d9b5ce2a2d06e101884667dbac077a | quote: There is no signup credit. - [4] 2026-11-03: an MCP server rents GPUs for an MCP client after approval | capture: sha256:78cd8c2ec45d24746a206dc7cf4c5a1881d9b5ce2a2d06e101884667dbac077a | quote: The CloudGPU MCP server lets Cursor, Cline, opencode, Claude Code and other MCP clients quote a price, rent a GPU once you approve - [5] 2026-11-03: the top-up bonus from 2026-11-03 is USD 10 for a single top-up of USD 100 or more | capture: sha256:b86e1ed2592ffe42b22061507014c0bfc5a8e41f6d9d73fb61b4b95f7b0c0bdf | quote: From 3 November 2026, a single top-up of $100 or more earns a $10 bonus, once per account - [5] 2026-11-03: capacity comes from third-party suppliers | capture: sha256:b86e1ed2592ffe42b22061507014c0bfc5a8e41f6d9d73fb61b4b95f7b0c0bdf | quote: Compute capacity is sourced from a pool of third-party data-centre and supplier networks. - [6] 2026-11-03: computeprices.com lists CloudGPU as a GPU rental provider with 11 GPU models, prices collected 2026-11-03 | capture: sha256:206e5c483427c9b2a37d593b2f441d1c95712908396509895a7de8247ef3eda0 | quote: CloudGPU offers GPU rentals and model APIs for various AI applications - sources: - [1] https://cloudgpu.app/pricing - [2] https://cloudgpu.app/ - [3] https://cloudgpu.app/payments - [4] https://cloudgpu.app/faq - [5] https://cloudgpu.app/terms - [6] https://computeprices.com/providers/cloudgpu