Start earning with the hardware you already own

Turn your idle PC into income — no upgrades, no effort.
tensor.spot does the work, you earn.

$395 – $585gaming PC $148 – $214laptop from 2018 $102 – $137office PC, no graphics card $0your own money
*All earnings shown are estimates, based on how average hardware performed over the last 30 days. No sign-up required. Payouts are made in US dollars.
1DownloadLightweight Windows client. No sign‑up required.
2RunLaunch the app to receive your access key — enter it on our website to unlock your dashboard and start earning.
3EarnGet paid in crypto, PayPal, or bank transfer.
tensor.spot paired with your dashboard Open my dashboard →
Earned today $94.10 1,284 tasks accepted today
This week$614.08
RendererNVIDIA GeForce RTX 3060
Graphics APIOpenGL 4.6 on D3D11
Memory held2.1 GB resident
GPU temp64°C, fan 38%
Uptime4h 12m
Active workloads
Training 46%Inference 31%Other 14%Rendering 9%
$88 Fri
$91 Sat
$79 Sun
$93 Mon
$86 Tue
$82 Wed
$94 Thu
Current job
Accepted today1,284
Token rate1.4 ktok/s
Average processing time128ms
boottensor.spot agent 1.4.0 — Windows 11
cpuAMD Ryzen 7 5800X8 × 3.80 GHz
gpuNVIDIA GeForce RTX 306012 GB
glOpenGL 4.6 on D3D11
memMemory budget: 12 GB VRAM + 32 GB RAM6 threads
benchbenchmark12.4 TFLOP/s
Each line is a computing task — hover to pause.
Live network activity

Join thousands already earning

Machines online49,934
Paid out last month$880,300
Paid out all time$128,501,311
Transactions per hour14,700
Machines working right now
Germanytu******n RTX 3060 Laptop
Türkiyesi******0 RX 7600
United Statesus******9 RTX 5060
Türkiyets*****5 RTX 3050
Franceca****2 RX 6750 XT
Canadaam******0 3 × RTX 4070 Laptop
Bahrainca***2 GTX 1650
South Koreaus******9 GTX 1660 Super
United Kingdomvi******e Ryzen 5, no graphics card
Cyprussw******t RX 7600
Japanor******5 3 × GTX 1070 Ti
Spainlu******i RTX 4060 Laptop
Belgiumne****g RTX 5080
United Kingdomgh******e RTX 4070
Polandbe******3 RTX 3060 Laptop
Türkiyeid*****h RTX 3080 Ti
Swedenvi******2 GTX 1050 Ti
Switzerlandtu******6 RTX 5060
Live transactions
Germanytu******n RTX 3060 Laptop | Embeddings credit +$0.11
Türkiyesi******0 RX 7600 | Fine-tuning payout, Venmo −$119
United Statesus******9 RTX 5060 | Text inference switch
Türkiyets*****5 RTX 3050 | Vision inference credit +$0.24
Franceca****2 RX 6750 XT | Vision inference credit +$0.27
Canadaam******0 3 × RTX 4070 Laptop | Text inference payout, SEPA transfer −$366

Estimate your earnings

Not sure what's inside your machine? Use auto-detection and we'll figure it out.

tensor.spot | income estimate *based on 30-day history

Your PC configuration select your components

handles the heavy workloads
1
manages smaller tasks
1
determines model size
when your PC is idle
Monthly earnings
$295 – $384
Graphics cardNVIDIA RTX 4070$250–$326
ProcessorAMD Ryzen 5 5600$45–$59
System memory32 GB-1% task size
Hours per day12h50% of the day
Memory budget12 GB VRAM + 32 GB RAM
Used by the job 9.8 GB Headroom, never swapped
Compatible workloads
EmbeddingsData labellingRenderingVision inferenceSpeechText inferenceVision trainingSimulationFine-tuningLanguage training

It runs itself

Autopilot$17.43 / day

Always on the best-paying job

Dozens of labs and studios post work at the same time, and the price of GPU time moves with demand. Every five minutes the program compares the open jobs and moves your card to the best-paid one. All you see is the number going up.

Open jobs$ per day
Vision training $17.43 running
Text inference $16.56 standby
Language training $15.83 standby
Fine-tuning $14.05 standby
switched to Vision training, 5% above the previous rate
Memory-first scheduling7,275

Old cards aren't slow — they're short on memory

AI jobs fail on weak cards for one reason: the batch won't fit in memory. The agent reads your VRAM and system RAM, sizes each task to fit with headroom, and streams the rest. A 6 GB GTX 1060 simply takes a smaller piece — slower per hour, but stable, and paid the same rate per unit of work.

Task size × batchstep 51
Best so far: task 2.1 GB, batch 8$20.57/day
Stays out of your way12 h/day

Works when you're away, pauses when you're back

Open a game, an editing timeline or just a video in the browser and it steps aside within a second, without slowing anything down. Walk away and it picks the work back up. You can also set the hours by hand — click the cells.

Automatically when idle
00:0006:0012:0018:0024:00
On this schedule, an RTX 4070 earns $571 – $744 a month (30-day average)
Stays cool80%

Quieter and cooler than a game

The program holds the card at 65–80% of its power: roughly a fifth less money for a markedly cooler, quieter machine.

50% — quieter100% — faster
69°CTemperature
59%Fan speed
90%Earnings
Why an old card still earns

The bottleneck is memory

No large AI model fits on a single machine — not on an RTX 4090, not on a data-centre card. Work arrives pre-split, and your task size is set by video memory plus the system RAM behind it.

Ordinary schedulers give every machine the same batch. When it doesn't fit, the card swaps to disk and the job fails — which is why old hardware gets written off. tensor.spot measures your memory first and never sends more than fits.

1
Read the memoryVRAM, system RAM
2
Size the taskFill about 80% and keep the rest as headroom
3
Adapt while it runsOpen a game and the next task shrinks — no crash, no lost work
The same job on three machines
GTX 1050 Ti — laptop4 GB VRAM
Working set 3.2 GB Headroom
RTX 3060 — desktop12 GB VRAM
Working set 9.8 GB Headroom
RTX 4090 — workstation24 GB VRAM
Working set 19.7 GB Headroom
Still not sure?

Try it for an evening. Nothing to lose

Leave the computer on for one evening and look at the number in the morning. Whatever the card earned is yours — withdraw it whenever you like, with no minimum and nothing to wait for. Not convinced? Uninstall it like any other program; nothing stays behind.

No money to put inNo sign-upNo wallet to set upNo AI knowledge neededNo need to leave it on all dayNo hardware to buy
Try it for one evening

Where the money goes

What your machine computes, who pays for it, what we keep, and how you get the rest.

1 — Your machineYour hardware does the workGraphics card and processor both take jobs — a training batch, an inference request, a render — while the machine is idle.
2 — The buyersCustomers pay by the minuteAI teams, studios and universities across Europe and the US. tensor.spot puts you on the best-paid job you qualify for.
3 — The split$98 of every $100 is yoursThat is the whole fee. Nothing else is deducted.
What you need

Almost anything that powers on qualifies

A Windows 10 or 11 desktop or laptop with an internet connection. What you can take on is decided by memory, not by model year: 4 GB of video memory is enough to start, 8 GB of system RAM keeps the streaming smooth, and a machine with no graphics card earns through the processor instead. Every task is cut to fit the hardware you actually have.

4 GB video memory8 GB system RAMNVIDIA GTX 10/16, RTXAMD RX 500 and newerLaptops from 2017Integrated graphics, Intel ArcProcessor-only machines
What you can check

Every dollar can be traced back

The app keeps a record of every action and every payout. Each payout carries a receipt — which jobs ran, for how long, at what rate — and you can open it on any device that has your key.

Action logA receipt for every payoutRate historyHours computed
What it does to your PC

No damage, no wear, no slowdown

Graphics card and processor are held at 65–80% of what they are rated for — lighter than a typical game — and temperature, fan speed and memory are read continuously. The app does not need your files or your passwords, and has no access to either.

Lighter than gaming loadSensors read continuouslyNo access to your filesUninstalls like any program
How you get paid

The balance moves as the work clears

Every accepted task is added to your balance straight away. Pick the service you already use — there is no minimum, and the money reaches your account the same day you ask for it. Customers are billed in dollars and you are paid in dollars.

In the United States
PayPalVisa / Mastercard debitACH bank transferCash AppVenmoZelleApple PayGoogle PayPayoneer
In Europe
SEPA transferRevolutWisePayPalVisa / MastercardBLIK (PL)iDEAL (NL)Bancontact (BE)Bizum (ES)Swish (SE)Papara (TR)Skrill
Everywhere else
SWIFT transferPayoneerWiseVirtual debit cardCrypto walletUSDT (TRC-20)USDCBitcoin
FAQ

Common questions

Short answers. If something is still unclear, a real person answers within the hour.

Is my computer good enough?

Almost certainly. What decides the size of your jobs is memory, not the year on the box: 4 GB of video memory is enough to start and 8 GB of system RAM keeps the streaming smooth. A laptop from 2017, an office machine with no graphics card, or even integrated graphics will all take work — the agent just hands them a smaller task, at the same rate per unit of work.

I don't know which graphics card I have

You don't need to. A web page cannot read your hardware, so the estimate on this site is only as good as the parts you picked. The agent reads the real thing — card, video memory, system RAM, and what else the machine is busy with — in about 60 seconds, and replaces the guess with a figure for your exact computer.

Will it damage my computer or slow it down?

The card runs at 65–80% of its power, lighter than a typical game, and the agent reads temperature, fan speed and memory every two seconds. It never touches your files or your passwords — it has no access to them. Sit down at the machine and the work stops within a second; walk away and it picks up again.

How and when do I get paid?

Every accepted task is added to your balance straight away — there is no payment cycle to wait for. Withdraw any amount at any time to PayPal, a Visa or Mastercard debit card, an ACH or SEPA transfer, Wise, Revolut, Zelle, Cash App, BLIK, iDEAL, Bizum, Swish, Papara, Skrill, Payoneer, or a crypto wallet in USDT, USDC or Bitcoin. No minimum, a 2% fee, and the money reaches your account the same day you ask for it.

What exactly is my computer working on?

Slices of AI work that labs, studios and universities pay for by the minute: a batch from a training run, an inference request, a few rendered frames, a set of embeddings. Your machine only ever sees encrypted tasks from a larger job — never the dataset behind it, never the finished model.

Do I need an account?

No. The first time the program runs it shows you a key that belongs to that machine. Paste that key into this site to open your dashboard. We store no email, no password and nothing that could identify you — and if you lose the key, the program shows it again.

Can I talk to a person?

Yes. Support runs on Telegram and answers within the hour between 9:00 and 23:00, in any of the languages this site is published in.

Support on Telegram
Still unsure whether it is worth it? Download for Windows (40 MB)
$0.00paid out to users last monthclosed month, settled in full — every payout carries a signed receipt
0machines online right now27% of them laptops — 10% cards older than 2019
$0.00average earned per machine, per dayacross every machine in the network — 30-day average
2 minutes to install, 60 seconds to check

See what your PC could earn

Download, press one button, and see your estimate — even on a six-year-old machine. Not for you? Uninstall it, nothing lost.

Download for Windows (40 MB)