A marketplace for idle GPUs

One side has GPUs sitting idle.
The other needs more compute.

Tom's gaming rig sleeps 16 hours a day. Maya is up at 2am paying Replicate $23 for a fine-tune that times out. Wattlend brings them together. Tom earns money from his idle GPU. Maya gets her fine-tune for $1.60. The platform takes 30%.

I need compute →I have a GPU →

No setup · No minimums · Pay only for what you used · Cancel anytime

How it works in 6 frames

A real walkthrough of what happens when a seller plugs in and a buyer rents. Plays automatically — pause anytime to read.

🧑‍💻

User A · Berlin

RTX 4090 · 24 GB

⚡ Wattlend

Agent connected

✓ Listed at €0.40/hr

Frame 1 of 6

1. User A plugs in their GPU.

A gamer in Berlin has an RTX 4090 that sits idle while she sleeps. She installs the Wattlend agent. The platform auto-detects her hardware, suggests a fair hourly rate, and lists the machine for rent.

How the marketplace works

Supply meets demand. We're the bridge.

Idle GPUs on one side. AI workloads on the other. Wattlend matches them, handles the plumbing, takes 30%. The seller earns money they wouldn't have. The buyer pays less than they would have. Same GPU. Better outcome for both.

Supply

Idle hardware looking for income

Gaming PC with an RTX 4090 (idle 16h/day)
Old crypto mining rigs (post-merge paperweights)
AI workstations used nights/weekends only
University lab GPUs idle off-hours
Small data centers / co-lo with spare slots

Earns $80–$1,200/mo on hardware that would otherwise sit idle.

$$ pays seller

locked hourly · 70% to seller

⚡ Wattlend

Matches buyers + sellers. Handles routing, billing, payouts. Takes 30%.

🖥 compute

OpenAI-compatible API · per-rental URL

Demand

Workloads that need more compute than the laptop has

Indie devs fine-tuning Llama-3 on side projects
AI startups burning AWS budget on inference
Image-gen bursts (1,000 SDXL renders, then nothing)
Audio transcription jobs (hours of podcasts)
Data scientists doing weekend analysis

Pays 60% less than AWS. No Docker, no SSH, no instance types — just "run my job."

No one in this picture would have used the GPU otherwise. Tom's 4090 was sleeping. Maya was about to give up on her side project. Wattlend's 70/30 split funds the platform — the buyer pays less than they would have, the seller earns money they wouldn't have, the platform takes a fair cut for matchmaking.

Real users, not personas

The four humans Wattlend is built for

Two on the demand side, two on the supply side. Real names, real moments, real reasons they pick Wattlend over the alternatives.

Real demand-side users — buyers

People who need compute they don't own.

DEMAND · BUYER

Maya

Indie AI dev · Austin · 28

Backend engineer by day, AI side projects by night. Last weekend she stayed up until 2am trying to fine-tune Llama-3-70B on a small dataset for a writing assistant. Replicate burned $23 without finishing. RunPod gave her three Docker compose errors. Colab kept disconnecting. OpenAI was off the table — her dataset is sensitive.

The moment

11pm Friday. Just watched a Replicate run time out and eat her budget. Googles "rent a GPU."

What Wattlend gives them

Click "Fine-tune" on the dashboard → autopilot picks a $0.40/hr 4090 → 4 hours = $1.60
No Docker, no SSH, no instance types, no AWS console
Her existing OpenAI SDK code works against the rental URL — no new framework to learn

Who they're NOT: An enterprise buyer with SLA requirements. We can't serve Maya's boss yet.

DEMAND · BUYER

Raj

AI startup founder · 3 employees · raised $500K seed

Building an AI customer-support tool. Burning $4,800/mo on AWS GPU instances. His CFO (who is also his wife) flagged it as 60% of their burn. Can't go full-OpenAI because their pitch is "private model running on your data." Needs cheaper inference for non-prod environments and occasional fine-tunes.

The moment

Q4 board meeting prep. Googles "cheaper than AWS GPU."

What Wattlend gives them

Same OpenAI-compatible API his app already calls — flip a base URL env var, save 60%
Per-rental endpoints means each environment (dev/staging) gets its own machine, billed only for hours used
Cancel anytime — no annual commit, no reserved instances

Who they're NOT: A Fortune 500. We don't have SOC 2, no enterprise SLAs, no dedicated support.

Real supply-side users — sellers

People with idle hardware to monetize.

SUPPLY · SELLER

Tom

Gamer · college student · 19

Paid $1,800 for an RTX 4090 mostly for Cyberpunk and Valorant. The card sits idle while he's in class (9–4) and asleep (12–7). He's seen Salad and NiceHash ads but mining isn't profitable post-merge and he doesn't want crypto on his tax return.

The moment

Finals week, broke, $300 power bill. Googles "make money from gaming PC."

What Wattlend gives them

Install agent → autopilot lists his GPU at the going market rate
Set quiet hours (4pm–midnight) so it never rents during his gaming time
Earnings show up. Last month: $80–$140 net after electricity. His GPU effectively pays for itself in ~12 months.

Who they're NOT: Someone running an actual business. Tom doesn't want phone calls about uptime.

SUPPLY · SELLER

Vinod

Ex-ETH miner · electrical engineer · Sacramento · 47

Ran an Ethereum mining op 2019–2022. Post-merge his 32 RTX 3080s and 3090s are paperweights. Talks to his old miner Discord every week — they're all asking what to do with the hardware. Some are listing on vast.ai but it's clunky for managing 30+ machines.

The moment

Every couple of months, Googles "list my mining rig for AI" hoping the answer has gotten better.

What Wattlend gives them

Bulk listing flow (Phase 2 of the rebuild) — manage 30+ machines from one dashboard
Hourly-locked earnings = predictable monthly income, $400–$1,200/mo for his fleet
Pro view shows fleet stats so he can run it like a side business

Who they're NOT: Someone who needs one-click setup. Vinod will RTFM, but he wants real fleet tools — and his Discord of 200+ ex-miners is paying attention.

Honest disclosure

Who Wattlend is not for

Most platforms in this category claim to serve everyone. We don't. If you're in one of these buckets, you'll be unhappy with us — please use the alternative we point you to.

Fortune 500 enterprise
Needs SLAs, SOC 2, dedicated support, multi-year contracts. We have none of that yet.
→ AWS, GCP, CoreWeave
Researchers running 6-week training jobs
Need stable hardware that won't drop mid-run. Community supply varies too much.
→ Lambda Labs, CoreWeave, university clusters
Real-time inference at million-request-per-day scale
Latency variance from community machines kills strict SLAs.
→ Together AI, Fireworks, Anyscale
People who want literal ChatGPT
ChatGPT is OpenAI's closed product. We can't redirect it. (Open-source ChatGPT clones work great on Wattlend, though.)
→ chat.openai.com
Anyone planning sketchy use
Sellers and the platform will report abuse. We coordinate with abuse@ and law enforcement.
→ nowhere good

For developers

Every rental gets a private OpenAI-compatible URL + Bearer token. Drop into any existing OpenAI SDK code by changing two lines. The model serving you is whatever the seller has installed (typically Llama 3 / Mistral).

curl https://api.wattlend.com/v1/rentals/$RENTAL_ID/chat/completions \
  -H "Authorization: Bearer $RENTAL_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "llama-3-8b-instruct",
    "messages": [{"role":"user","content":"Hello from a rented GPU."}]
  }'

Or use Open WebUI / LibreChat / Cursor / Continue / Chatbox — every rental's detail page has step-by-step setup for each. Same Bearer token, same URL.

If you're Maya or Raj

Rent a GPU. Use it from your laptop.

$0.30–$2/hr depending on hardware tier. Locked at booking. Cancel anytime.

Try it — start with $5 →

If you're Tom or Vinod

List your idle hardware. Earn while you sleep.

Install agent. Autopilot prices it. Set quiet hours. Get paid weekly via Stripe.

Start listing →