AI is here to stay, it seems, and a lot of people have leaned into the change rather than resisting it. Many people engage with AI over their web browser since major AI services are available there, thereby bypassing the need to have a strong PC that’s optimized to run AI. However, other folks want to run AI directly on their PCs. That is a different category entirely.
To run AI, you need a pretty strong computer. You can probably run something simple like Nvidia’s ChatRTX without a ton of resources, but if you’re generating images, video, or doing development work, you probably want more than just the bare minimum. So, if you’re getting ready to build an AI, here are some tips on how to do it.
Pick your operating system
An AI-powered PC needs to run an operating system that can actually use AI. All three major computer operating systems can do this, including Windows, MacOS, and Linux. Generally speaking, Ubuntu and Windows are the two most popular operating systems for running AI, and they should be at the top of your list if you’re building an AI-powered PC.
Windows is a little better for basic stuff because Copilot is built in. Copilot is in several of Microsoft's apps and in the OS itself, plus you can download the native Copilot app from the Microsoft Store. Ubuntu is superior if you want to build an AI PC from scratch, as all the tools work there and you have the freedom to set things up the way you want. However, Ubuntu is harder to learn if you’ve never used it before.
Let’s talk specs
An AI-powered PC requires some hefty specs if you want your AI to run well. That includes a large amount of RAM and storage. The more you have of both, the better your AI will run. If you’re wondering why these two things are so expensive right now, this is why. AI data centers are eating up all the RAM and storage so each company’s AI runs better.
The baseline you want to shoot for is about 32GB of RAM and around 1TB of storage, although the more you have, the better things will be. If you skimp here, you won’t have enough space to download AI models or enough RAM to run them.
An Nvidia GPU
I am a fan of AMD GPUs myself, but if you want a PC to run AI, then Nvidia is the way to go. Many of the biggest tools have direct compatibility with Nvidia, and run better when compared to similar AMD GPUs. You want the highest-end card you can afford with the most possible VRAM. Most guides have a floor at 8GB of VRAM, but there are AI models that can easily make use of the 32GB of VRAM in the exceedingly expensive RTX 5090, so there’s no ceiling here.
A good start is the Nvidia RTX 5070, which comes with 12GB of VRAM, but if you can swing for the stronger RTX 5080 or RTX 5090, you’ll be much better off.
A CPU with an NPU
The CPU is still important, just not quite as much as the GPU, RAM, and VRAM. Modern CPUs have what is called an NPU, which stands for neural processing unit. These are used specifically for AI tasks, and having a PC with one sets you up for success. All of the major chipmakers have this, including the Intel Core Ultra series, the AMD Ryzen AI series, and even the Snapdragon X series, although you won’t be making a PC with that last one. Pick whichever one you like most.
All of this applies to laptops too
All of the above advice applies to people who are building a PC, but they also apply to people shopping for laptops. Look for ones with a high-end Nvidia GPU with plenty of VRAM and a CPU that includes an NPU. Most laptops that fulfill these qualifications are user-upgradeable, so you can always add more RAM and storage later, although if you can find one with 32GB of RAM and at least 1TB of storage, you’ll be pretty well set for a while.