Short Overview Guide

AI Rental Rig Profit Overview: $3,000 RTX 4090 Build on Salad

This short guide summarizes Modern Mining's 2026 AI rental rig video: what was built, why the RTX 4090 and RAM matter most, what setup choices affected Salad earnings, and how the rough profit math works.

1

Overview

The video builds a dedicated AI compute rental PC intended to earn passive income on Salad. The creator's goal was not to build a balanced gaming PC; it was to optimize for components that affect rental demand and earnings.

Main idea: spend money where the rental platform pays for it: GPU performance/VRAM and enough system RAM. Keep the CPU, motherboard, frame, and other parts practical rather than premium.
$3,000

Approximate build cost target in the video.

RTX 4090

Primary earning component, valued for 24 GB VRAM and strong AI demand.

~$71/mo

Creator's rough profit estimate after fees, power, and uptime adjustment.

2

Build snapshot

The sample rig used these kinds of parts:

  • GPU: used ASUS TUF RTX 4090, bought locally for about $2,100.
  • Power: 1000 W PSU with a direct 12V-2x6/12VHPWR-style GPU cable.
  • Motherboard: budget B550 board.
  • CPU: Ryzen 5500, with the note that CPU performance is not the main revenue driver.
  • RAM: 64 GB DDR4 in the build, with the creator emphasizing that RAM matters for these rigs.
  • Storage: 1 TB NVMe SSD.
  • Monitoring/control: smart plug for wattage tracking and remote power cycling.
  • Frame/cooling: open frame and simple air cooling.
Buying used GPUs is risky: the creator recommends proof of the card working, caution with cash transactions, and comparing market prices before buying.
3

Setup highlights

  1. Assemble and boot-test the rig. Confirm the GPU, RAM, SSD, motherboard, and CPU all appear correctly.
  2. BIOS: restore after AC power loss. Set the rig to power back on after power is restored so a smart plug can act as a remote reboot tool.
  3. BIOS: enable virtualization. On the MSI board shown, this appeared as SVM Mode. Salad needs virtualization support for best compatibility.
  4. Skip XMP if stability matters more than RAM speed. The creator did not see earnings benefits from faster RAM settings and preferred stability.
  5. Install Windows, NVIDIA drivers, and Salad. Current NVIDIA drivers are necessary; outdated drivers can block earnings.
  6. Configure Salad manually. The video recommends manual setup, elevated permissions, disabling crypto mining, and adjusting content/video settings based on the creator's experience.
  7. Add remote management. Chrome Remote Desktop plus a smart plug gives two ways to recover a hung rig.
  8. Measure power. Use a wattage-monitoring plug to convert revenue into actual profit.
4

Profit math from the video

The creator's rough model used these assumptions for an RTX 4090 job on Salad:

Revenue assumptions

  • Gross earning rate: $0.18/hour.
  • Gross daily revenue at 24 hours: $4.32/day.
  • Fee assumption: 10%.
  • Job availability adjustment: 80%.

Power assumptions

  • Measured draw: roughly 450-466 W while running.
  • Video math uses about $0.04/hour.
  • Daily power cost estimate: $0.96/day.
  • Heat may be useful in winter but costly in summer if air conditioning offsets it.
Gross daily revenue:         $4.32
After 10% fee:               $3.89
Minus estimated power/day:   $0.96
Approx. profit/day:          $2.93
80% uptime-adjusted year:    ~$855/year
Monthly equivalent:          ~$71/month
Simple $3,000 payback:       ~3.5 years
Do not treat this as guaranteed income: GPU rental demand, Salad customer demand, power rates, platform policies, and used hardware prices can all change.
5

Reality checks before copying the build

  • Existing hardware is the safer starting point. The video frames Salad as especially attractive for people who already own capable GPUs.
  • High VRAM matters. The creator reports more reliable demand for top-tier GPUs like 4090/5090 than for lower-end cards.
  • Internet and upload speed matter. The creator suggests strong broadband and notes upload speed is important.
  • Heat management matters at scale. Multiple rigs can meaningfully heat a room; exhausting heat outside can avoid doubling costs through air conditioning.
  • Resale value is part of the creator's investment thesis. The creator treats hardware resale value as a release valve, but that depends on market conditions and buying at good prices.
6

Risks and caveats

  • Used GPU failure or scams: the GPU is the largest cost and biggest single-point financial risk.
  • Demand fluctuation: a GPU may not have a job 24/7, and lower-end cards can see less consistent demand.
  • Power cost variance: the math changes quickly if your electricity price is higher than the creator's estimate.
  • Platform fees and rules: cash-out fees, supported workloads, settings, and rewards can change.
  • Windows interruptions: updates, driver issues, or reboots can interrupt jobs.
  • Not financial advice: the source video itself includes a disclaimer; treat this as a hobby/experiment overview, not an investment recommendation.
7

Sources

Note: Several source links in the video description are affiliate links. This guide preserves them as source context and does not endorse a purchase.