Private beta — listings reviewed by hand

Serious GPU power for AI image and video, at half the going rate

Describe what you need: workflow, VRAM, hours per month. We review every request by hand and match it with a host. No account and no payment on this page.

Post a request

Your request appears on the live board only after we approve it. Nothing is public before review.

Why we think we can be cheaper

Our supply comes from idle consumer GPUs in China, where electricity costs a fraction of what hosts pay in the US or Europe. We think that cost gap is what lets us price lower. It is a hypothesis the beta has to prove, not a margin we have already achieved.

Two tiers, pick your trade-off

Instant

$0.25/GPU-hour
−49% vs Vast.ai median

Aimed at hosts with a public IP and stable uptime. Matched by hand, so start time depends on the supply available at the time.

Queued

$0.15/GPU-hour
−69% vs Vast.ai median

Runs on idle consumer machines. May be interrupted and resumed. Best for batch work that can wait.

RTX 4090 pricing, compared

ProviderLowestMedian
Vast.ai$0.14$0.49
Salad$0.16—
RunPod$0.34—
Our target$0.15$0.25

Competitor figures are live listings captured 2026-09-16. We compare against the median, not the cheapest single listing — the floor price is one machine, not a market.

The electricity gap

RegionElectricityCost per GPU-hour
China$0.08 / kWh$0.05
United States$0.17 / kWh$0.10
Europe (avg)$0.25 / kWh$0.15
United Kingdom$0.33 / kWh$0.20
Germany$0.37 / kWh$0.22

Full-system draw for an RTX 4090 rig is about 600W. This is why listings in Europe cannot go much below their current floor.

How we handle your material

Isolation is a launch gate

We will not accept jobs until they run in isolated containers that are destroyed after each run. This is not live yet, and we will say so when it is.

No payment on this page

Posting a request costs nothing and creates no obligation. Payment terms are agreed when we match you with a host.

Nothing confidential, for now

During the beta we recommend general-purpose generation only. Do not send unreleased client material yet.

Post your GPU need

About three minutes. Contact details stay private; only the headline and specs can appear on the board.

Basics

6–80 characters. No contact details, no names, no URLs.
Visible only to the platform team. It is never shown on the board.

Contact

Overseas users can use WhatsApp, Telegram or Instagram. Mainland China users can use WeChat or phone. Contact details are used only to match you and are never shown on the board.

Email is always required (for sign-in). Pick the channel we should use to reach you first.

What you need

For example: ComfyUI + an SDXL checkpoint, batch renders at night.
Preferred GPUs (optional)
Leave blank if any card works.
Job style

Tier and budget

Tiers you accept
Our instant tier is $0.25; queued is $0.15.

Data and payment

Data sensitivity
Not accepted during the beta.

1 USD ≈ 7.1 CNY

For GPU owners

Sitting on an idle GPU?

We plan to rent it to studios that need compute. Your main cost is the electricity it already uses. Hosts are onboarded by hand during the beta.