TapGPU: Hourly GPU Rental with App Templates
TapGPU rents GPUs by the hour to AI developers and creators — consumer cards like the RTX 4090 and 5090 alongside A100, H100 and H200 — with popular AI apps preinstalled as one-click templates. It is a China-market service: prices in CNY, payment by Alipay, site published in Chinese. This profile describes how it works and deliberately quotes no prices, for reasons explained below.
How this listing got here. TapGPU submitted this listing through our provider listing form. Everything below is drawn from TapGPU's own website (homepage, checked on November 9, 2026) and its listing application. On prices: the application's quoted range and a third-party snapshot of the same platform disagreed with each other, and TapGPU itself says availability and price are whatever the console shows at the time — so this profile publishes no rate figures at all. The live console on tapgpu.com is the only source of truth.
On this page
TapGPU at a glance
| Item | What TapGPU publishes |
|---|---|
| Service | Hourly cloud GPU rental for AI training, inference and generative workloads |
| Consumer GPUs | RTX 3090, RTX 4090, RTX 5090 (listed for image generation, inference and small-to-mid training) |
| Data-center GPUs | A100, H100, H200, L40S, A40 (listed for large-model training and high-memory tasks) |
| Professional GPUs | RTX A6000, RTX 6000 Ada, RTX PRO 6000 |
| Configurations | 1, 2, 4 or 8 GPUs per instance, subject to available machines |
| Billing | Hourly by actual runtime, priced in CNY (a per-minute equivalent is also shown); monthly whole-machine rental also offered |
| Payment | Alipay top-up of an account balance — the only payment method its FAQ lists |
| App templates | Preinstalled environments including ComfyUI, Stable Diffusion WebUI, Kohya_ss, Text Generation WebUI and PyTorch (see list) |
| Access & storage | Browser Jupyter/terminal and SSH; system disk plus attachable data disk; a shared model disk reusable across instances |
| Prices | Not quoted in this profile — live console only (see caveats) |
| Contact | Customer service inside the console (its homepage points to an in-console service channel and a contact page); no public support email found on the pages checked |
| Official site | tapgpu.com (published in Chinese) |
Four things to know before you buy
Each point below is stated in TapGPU's own pages or follows directly from them. They matter more than any rate card when you compare TapGPU with the providers in our directory:
- There is no price figure we can honestly print. TapGPU's application quoted an indicative range of ¥0.58–4.04 per hour across its catalogue (as of November 9, 2026); a third-party snapshot of the same platform showed ¥0.77–3.12 per hour. Both cannot be the going rate, and the platform's own pages say the models, counts and prices actually available are those shown in the console at the time. So this profile quotes neither figure as fact — check the console, per GPU model, before budgeting.
- Payment and language mark this as a China-market service. Its FAQ lists Alipay as the payment method, prices are in CNY, and the site is published in Chinese. If you cannot pay by Alipay or read the console, treat that as a practical blocker, not a detail.
- Stopped instances can keep costing money. TapGPU's own FAQ warns that after an instance stops, its system and data disks may continue to incur storage charges, and advises deleting storage you no longer need. Data on a disk survives until the disk is deleted — and is unrecoverable afterwards — so the delete decision cuts both ways.
- The operator is not reachable through the application it sent us. The email address supplied with TapGPU's listing application sits on an
.invaliddomain — a reserved, undeliverable address — and the pages we checked publish no operating company name, address or support email. Support is offered through the in-console customer-service channel. Do your own diligence on who you are paying before topping up a balance.
What you can run: app templates
TapGPU's pitch to creators is that common AI apps come preinstalled as templates you select at launch, with no manual environment setup. As listed on its homepage:
- Image generation — Stable Diffusion WebUI (Automatic1111 / Forge), ComfyUI, Fooocus, InvokeAI, FaceFusion
- LoRA training & fine-tuning — Kohya_ss, AI Toolkit, FluxGym, Diffusion Pipe
- Video generation — Wan2GP, InfiniteTalk, Advanced Live Portrait
- Audio generation — IndexTTS2, RVC, TTS WebUI
- General — Text Generation WebUI, a GPU cloud desktop, and PyTorch deep-learning environments
The workflow its homepage describes: register and top up by Alipay, filter instances by GPU model and count, pick a template, review the estimated cost, and launch; when the instance is ready you open the app's web page or a Jupyter terminal, and stop or delete the instance when done.
How billing works
Billing is hourly against actual instance runtime, deducted from an account balance topped up by Alipay; prices are displayed in CNY per hour with a per-minute equivalent shown alongside. An estimate covering the GPU, system disk and data disk is shown before you launch. A monthly option rents a whole machine for longer training runs. Eligible new users are offered a "newcomer price" on their first instance launch. Two cost details from TapGPU's own FAQ are worth repeating: storage can keep billing after an instance stops (see the caveats), and the price that applies is the console's live price for that model and configuration — which is exactly why this profile prints no figures, and why our pricing guide recommends screenshotting the console quote before any top-up.
Who runs TapGPU
The pages we checked (homepage, pricing, about and contact routes listed in its sitemap) publish the platform and its terms, but no operating company name, registration, headquarters address or team that we could verify, and no public support email. The contact email supplied with its listing application is undeliverable (an .invalid domain), so we have been unable to reach the operator to confirm these details, and no listing notification has been sent. If you represent TapGPU, contact us through our listing page with a reachable address and we will update this profile.
Comparing providers? Return to the GPU cloud provider directory for the official pricing pages of all listed clouds, read how GPU cloud pricing works, and sanity-check any quote in our GPU cost calculator.