AI News
TechCrunchAnthropic just launched tools to show companies what their AI spending is buyingTechCrunchWiz’s CTO on PHP2Shell and the coming ‘vulnpocalypse’TechCrunchThe ‘Vulnpocalypse’ is coming for the internet’s aging code, says Wiz’s CTO Ars TechnicaGoogle announces Gemini 4 Carbon, its latest frontier AI model that you can't useThe VergeMicrosoft is ending Copilot support on smart TVsArs TechnicaGoogle AI gets cagey when asked who won the 2020 electionsTechCrunchAs AI agents reshape the web, Google opens Gemini and Search to more appsTechCrunchChrome’s AI-powered Autofill can now dig through your Gmail to fill out formsThe VergeChrome’s enhanced autofill can now pull info from Gmail to help you fill out formsThe VergeMeta says it’ll probably delay its controversial new business terms againTechCrunchAI execs may be too out of touch to build for the general publicTechCrunchSouth Korea’s financial industry hit by fresh wave of cyberattacksTechCrunchVoice AI in vogue: Nvidia releases new open transcription modelThe VergexAI is buying out homeowners who promise not to sue over data center noiseTechCrunch7 days until TechCrunch Founder Summit: Save up to $100 on your pass before doors openThe VergeVoters in Texas agree: They hate data centers

Running AI on Your Own PC: What Hardware You Need

October 11, 2026 · AI Mind Rush

Most AI tools run in a browser and need nothing special. Running models on your own computer is different. It is private, has no monthly fee and works offline, but it depends heavily on one part: the graphics card.

Disclosure: our gear page contains Amazon affiliate links. As an Amazon Associate we earn from qualifying purchases.

Graphics memory is the limit

AI models load into the graphics card’s own memory, called VRAM. If a model does not fit, it either will not run or runs very slowly. As a rough guide:

  • 8 GB: small language models and basic image generation.
  • 12 to 16 GB: comfortable for image generation and mid-sized language models.
  • 24 GB or more: larger models and some video generation.

Speed matters less than memory. An older card with more VRAM is often more useful than a newer one with less.

NVIDIA or AMD?

Most AI software is written for NVIDIA cards first, so they are the easiest to set up. AMD cards can do the same work and often cost less for the same memory, but expect more setup and the occasional tool that does not support them. We run an AMD card ourselves and it works, with some patience.

System memory and storage

  • RAM: 32 GB is a sensible floor. Models that overflow the graphics card spill into RAM.
  • Storage: models are large files, often several gigabytes each. A fast NVMe drive makes loading quick, and a large hard drive is cheap space for the ones you are not using.

Do you need a powerful processor?

Not especially. A mid-range processor from the last few years is fine. Spend the budget on the graphics card.

A second, small machine

Automation tools such as n8n do not need a graphics card at all. A small, low-power mini PC can run them around the clock for very little electricity, leaving your main PC free.

Free software to start with

See the parts we mean on the AI gear page.

This post may contain affiliate links. We may earn a commission at no extra cost to you. Learn more

Get the AI Mind Rush weekly

New AI tools + the week's AI news in 3 minutes. Free cheat sheet when you join.

Free. Unsubscribe anytime. See our Privacy Policy.

Leave a Reply

Your email address will not be published. Required fields are marked *