Guide · Local AI

Inside PewDiePie's home AI server (and how to build your own)

By Tiny · how I pick

Updated October 2026. The specs below come from public reporting on his build. I keep prices out of the picks because they change every week.

As an Amazon Associate I earn from qualifying purchases. Buying through these links supports the site at no extra cost to you. Full disclosure

In late 2025, YouTuber PewDiePie (Felix Kjellberg) went all-in on running AI himself. He switched to Linux and built a huge home server with ten graphics cards and about 424 GB of VRAM, so he could run open-source models privately, with no cloud and no subscription. It reportedly cost somewhere between $20,000 and $41,000.

You don't need any of that to do the same thing. The software he uses is free and runs on a single cheap graphics card. His rig is the far end of a hobby anyone can start for a few hundred dollars.

Words you'll see in this guide
  • VRAM: the memory on your graphics card. AI models load into it, so more VRAM means you can run bigger models.
  • Self-hosting: running an app on your own hardware at home instead of paying a company to run it for you.
  • Open source: software whose code is public, so anyone can check it, use it, and usually run it for free.
  • Parameters (70B, 235B): a rough measure of a model's size. "8B" means 8 billion. Bigger models are usually smarter and need more memory.
  • PSU: the power supply, the box inside the case that turns wall power into power every part can use.
  • Docker: a tool that runs each app in its own sealed box, so apps don't break each other or your system.

See the full glossary

What's in his rig

Going by public reporting, this is a small AI workstation, not a gaming PC:

PartWhat he used
Graphics cards8 RTX 4090s modified to 48 GB each, plus 2 RTX 4000 Ada (10 total)
Total VRAMAbout 424 GB
Processor64-core AMD Threadripper PRO 7985WX
System RAM512 GB
PowerTwo power supplies of about 1,300 W each
Reported costAbout $20,000 (his figure), about $41,000 (a teardown estimate)

The 48 GB RTX 4090s are specially modified cards. Regular store-bought 4090s come with 24 GB. All that VRAM lets him fit very large models entirely in graphics card memory, which is what decides everything in local AI.

Why he did it

His reasons are the same ones behind this whole hobby: independence and privacy. Instead of renting AI from a company, he runs it on hardware he owns, so his prompts never leave the house.

The part that matters for you: the expensive part is the hardware, and you don't need most of it. The software, whether that's Odysseus or the popular Ollama and Open WebUI setup, is free and runs on a modest PC. VRAM only decides how big a model you can load. It doesn't decide whether you can start.

How to build your own, at any budget

You can head in his direction at a sane price. Here are three realistic starting points. For a full parts walkthrough, see the home AI server build guide.

Graphics card prices jumped this year. As of October 2026, the RTX 5060 Ti and used RTX 3090 cost far more than they did last winter, so check current prices. My GPU guide has the details.

1
Budget local AI build
Just start

Used tower + used RTX 3060 12GB

A PC you already have plus a used RTX 3060 12GB runs 7B and 8B models comfortably, and most 13B and 14B models too. That's useful for a private assistant, and enough to run Odysseus or Open WebUI. This is where almost everyone should start.

Check price on Amazon
2
Mid-range local AI build
The sweet spot

RTX 5060 Ti 16GB or used RTX 3090 24GB

A new RTX 5060 Ti 16GB or a used RTX 3090 24GB, with 32 GB of system RAM, runs 14B models with long conversations and gets close to 30B-class models. You get a fraction of his VRAM and most of the everyday usefulness.

Check price on Amazon
3
High-end local AI build
Heading his way

Two used RTX 3090s (or one RTX 4090), 64GB+ RAM

The high end for a hobbyist. Two used RTX 3090s pool to 48 GB and run a 70B model, the same class PewDiePie runs, for a small fraction of the cost. A single RTX 4090 24GB is simpler, if you can find one. This is as far as most enthusiasts ever need to go.

Check price on Amazon

The rest of the parts

Apart from the graphics card, it's an ordinary PC, so don't overspend. More detail in the build guide.

The free software that runs it

This part costs nothing. You have two good options:

Both run in Docker on Linux, and both keep every prompt on your machine. Before you buy anything, put your current PC into the Server Calculator. It might already run them.

What to take from it

PewDiePie's 424 GB, five-figure rig shows how far running AI at home can go. You don't have to spend $40,000 to get the benefit. A hobbyist can now run a private AI at home with no subscription, starting with a used $250 graphics card and free software.

Start with the sweet-spot build, buy for VRAM first, and grow from there. If you're weighing this against paying for the cloud, read local AI vs Claude and ChatGPT first.

Quick answers

What is in PewDiePie's AI server?

Reporting on his build describes ten graphics cards, eight RTX 4090s modified to 48GB each plus two RTX 4000 Ada cards, for roughly 424GB of VRAM. It pairs them with a 64-core AMD Threadripper PRO 7985WX, 512GB of RAM, and two power supplies of about 1,300 watts each. He put the cost near $20,000, and one detailed breakdown estimated closer to $41,000.

What software does PewDiePie use for his AI?

He built a custom web interface called ChatOS to run open-source models locally and have several chatbots work together and vote on the best answer. He later released Odysseus, a free, open-source, self-hosted AI workspace under the AGPL-3.0 license, which runs best on Linux.

Do I need hardware like PewDiePie's to run local AI?

No. His machine is an extreme showcase. The same free software runs on ordinary hardware: a single 12GB graphics card runs capable 7B to 14B models. VRAM decides which model sizes you can run, so buy what you can justify and grow later.

Related guides

Affiliate disclosure: Some links on this page are affiliate links. If you buy through them, I may earn a small commission at no extra cost to you. As an Amazon Associate I earn from qualifying purchases. I'm not affiliated with PewDiePie. This is my own write-up based on public reporting.