AI is everywhere. AI companies like OpenAI and Anthropic went from unknown to household names in just a few short years. Big Tech companies like Google, Meta, Amazon, and Microsoft keep rolling out their AI models across every product, from search to chatbot assistants and more. And, naturally, companies that manufacture computer chips and GPUs like Nvidia are raking in billions as they create the hardware that powers it all.
But what if consumers don’t want to send their data to Big Tech companies or fork over hundreds, if not thousands, of dollars per month to Big AI companies for tokens? While computer hardware companies like Nvidia may be struggling to keep up with demand, everyday users can still find everything they need to build out their own machine – an AI PC – and run AI models locally
But what exactly is an AI PC? What do you need if you want to build your own, or what should you look for if buying an AI PC pre-made? And how does it differ from other powerful computers built for specific purposes, such as gaming PCs?
To answer these questions, Mashable spoke with an expert who builds PCs for a living, Quoted Tech co-founder and CEO Kevin Jia. Founded in 2023, Quoted Tech is a Canada-based company that builds custom PCs for thousands of customers. Jia and company also provided Mashable with access to one of their prebuilt models, the Quoted One Pro Plus, an entry-level custom prebuilt made-for-AI computer, so we could test out exactly what makes an AI PC an AI PC.
What is an AI PC?
An AI PC is exactly what it says it is: A computer that is built to handle artificial intelligence, machine learning, and large language models (LLMs).
Technically, every PC can utilize AI nowadays, just like one can play games or edit videos on any computer, no gaming PC or video editing workstation required. Just as a gaming PC delivers peak performance and the best gaming experience, an AI PC is built specifically to run AI models.
As Jia pointed out, though, an AI PC doesn’t mean it’s a single-purpose machine, as many of the specifications you’d find in an AI PC can be found in a gaming rig or PC built for video processing, too.
« Users don’t need to choose between an ‘AI’ rig and a ‘gaming’ rig, » Jia explained. « An AI-oriented computer doesn’t have to be a single-purpose machine. It is a workstation that’s fully capable of AI work, but it’s also built to comfortably handle gaming and other professional workloads in one system. »
Could I just use my laptop for AI?
Of course, you can use ChatGPT, Claude, Copilot, and Gemini all on your $300 laptop. Cloud models can be run on any machine.
However, if you’re trying to run local AI processes, you’re going to need a PC built for the task.
« AI needs a lot of GPU processing power, and you need a lot of VRAM, and you need a lot of memory, and you need a decent CPU, and you need to be able to cool all of that in a decent tower, » Jia told us. « We’ve had customers come to us saying, ‘Hey, I bought this laptop for AI, and it doesn’t work, it just overheats, and it runs slow.' »
To get an idea of what’s inside a consumer-level AI PC, here are the specs for Quoted One Pro Plus that we tested out for this piece:
-
CPU: Intel Core i5-14600K (14 cores / 20 threads, up to 5.3 GHz)
-
GPU: NVIDIA RTX 5060 (8GB GDDR7)
-
Memory: 32GB DDR5 (2×16GB) Kingston Fury Beast @ 5600 MHz
-
Storage: 1TB Kingston KC3000 NVMe M.2 SSD (up to 7,000 MB/s read speeds)
-
Motherboard: MSI PRO Z790-VC WiFi
-
CPU Cooling: Thermalright Peerless Assassin 120 (white, dual-tower air cooler)
-
Power Supply: MSI MAG A650BN 650W, 80+ Bronze (non-modular)
-
Case: Fractal Design North (white with tempered glass side panel)
-
Operating System: Windows 11 Home
What’s the main difference between a PC built for AI and other PC builds?
Traditionally, when one thinks of a high-end PC, at least at the consumer level, they think of a tricked-out gaming PC.
While you can game on an AI PC and vice versa, custom PC builders like Quoted Tech know there’s a difference when building a PC for a specific purpose like AI.
One major difference? The emphasis on cooling.
« You can see that there’s a ton of airflow, » Jia said. « That’s going to be a huge non-negotiable for AI-built systems. »
After spending some time testing out the Quoted One Pro Plus, a « ton of airflow » actually seems like an understatement. For its line of AI computers, Quoted Tech uses Fractal Design’s North PC case.
The first thing you’ll notice is that the PC tower is much larger than you’d expect. Of course, they could make it smaller. This isn’t a technological limitation. The PC case is big on purpose. North was designed for maximally optimal airflow to cool down your PC.
« A lot of prebuilt systems out there are designed with a smaller form factor because they’re cheaper, » Jia says. « We’ve had people kind of come to us and say, ‘I’m running this on a pre-built gaming system, and it’s just crashing or slowing down, or it’s not performing.’ And we tell them how they have the equivalent of a large laptop. That system wasn’t designed to be pushed like this. »
You can feel a light breeze emanating from the Quoted One Pro Plus when it’s running. It’s like night and day when compared to my own, much smaller gaming PC. My PC tower could heat up my room if needed during the winter while playing a couple of minutes of Fortnite. In contrast, the airflow felt coming out of Quoted One Pro Plus could cool a small room during the summer.
« When you’re gaming, I would hope that most of us are only gaming for a few hours at a time, » Jia explained. « A lot of gaming systems, while they’re good for airflow, might not be top tier because they weren’t designed to be run for 7 or 8 or even 12 hours. »
« Whereas with AI, if you’re doing LLM work or anything like that, you could be running this thing for 2 or 3 days. » he continued. « You need stability. You need workstation-class cooling. You need workstation-class space. Those you cannot trade away. »
There also tends to be no limit to how much processing power you can fit into an AI PC. For example, Quoted Tech has built workstations that can host up to 8 GPUs simultaneously.
« That would be an awful gaming PC because most games won’t be able to utilize more than one GPU effectively, » Jie said. « If you said, ‘I’m gonna buy a Blackwell card to play computer games,’ I would say I think we need to rethink what you’re spending the money on because you could probably get 99 percent of the performance [in a gaming PC] for one-third of the budget. »
Why an AI PC over Apple’s Silicon-powered Macs?
In November 2020, Apple released its first line of Silicon Macs with the new M1 MacBook Air, MacBook Pro, and Mac mini. These Macs provided consumers with options more powerful than similarly priced laptop and desktop PCs.
As Apple has packed more and more processing power in these Silicon chips over the years, they’ve become popular with users looking to run local AI processes such as the AI assistant OpenClaw.
While these Macs remain good options for certain use cases, Jia explains why PCs remain a top choice for AI.
« Generally, most LLM and AI developers today build on NVIDIA’s CUDA language. That’s really the defining difference, » Jia said, referring to Nvidia’s proprietary Compute Unified Device Architecture, which requires Nvidia’s GPUs and thus can’t be run on Macs.
« Macs and Radeon-based systems simply don’t have the same cohesive CUDA-driven ecosystem for developers to work within, » he continued. « For anyone building or experimenting seriously with AI models, that ecosystem matters, it determines what tools you can use, what workflows you can optimize, and how easily you can scale your work. »
And, of course, PCs can be continuously upgraded with new hardware whereas consumers cannot upgrade the internals of their Macs.
To compare the two, I ran a generative AI model for images on both. Both completed the tasks based on the assigned prompts. This wasn’t surprising. After all, both were built to handle AI processing.
While the consumer-level hardware I had available wasn’t the best way to showcase the difference between PCs and Macs when it comes to processing AI, I did take advantage of having access to my own Mac mini M4 Pro and Quoted Tech’s Quoted One Pro Plus. To compare the two, I ran a generative AI model for images on both. Both completed the tasks based on the assigned prompts. This wasn’t surprising. After all, both were built to handle AI processing.
However, the difference in the heat emanating from the Mac mini and the cool air continuing to flow through the Quoted One Pro Plus was hard to miss. It’s easy to see how the form factor of a property-built AI PC will give it an advantage over the Mac in terms of performance.
How much is an AI PC going to cost me?
Obviously, there’s quite a wide range of pricing, as AI PCs can be built for consumer use or for large enterprise use. Jia informed me that Quoted Tech has built workstations that were priced as high as six figures. But individual users looking to run local AI models obviously don’t need to spend anywhere near that much.
Quoted Tech says that those looking to « dabble » in running local AI models could get started on an entry-level AI PC, like the Quoted One Pro Plus, for between $1,500 and $2,500.
