- A new startup has built an AI PC which has sold out of its first batch
- Ghost's Core is a 'personal AI computer that runs models entirely on device'
- It connects to your apps, services and devices, so it 'builds context on your life over time, and works continuously to help you'
Ghost is a startup in the AI field that wants to capitalize on the desire to have agents running locally rather than using AI in the cloud — and to get this done with a minimum of fuss.
TechCrunch reports that Ghost's first product is a PC for using local AI agents which runs those LLMs (Large Language Models) exclusively on-device, ensuring your data stays private (I'll come back to that point later).
The device is called Core, and Ghost explains on the product's site: "Core is a personal AI computer that runs models entirely on device. It connects to your apps, files, and devices, builds context on your life over time, and works continuously to help you, not just when you ask."
It is, if you will, a personal AI assistant on your desk at home, or as Zain Javaid, the founder and CEO of Ghost — which just raised $11 million in funding — put it, a "brain in a box".
Core is a PC which is an unassuming black box built around an Nvidia RTX Pro 4000 (Blackwell) graphics card with 24GB of video RAM, the central muscle for driving the AI, along with 64GB of DDR5 system RAM. The rated AI performance is 770 TOPS.
The PC comes with three AI models pre-installed: Qwen-3.8-Next, Qwen-3.8-27B, and Gemma-4-31B.
You can use your own model on the device, as you might expect, or any alternative for that matter, but the real selling point of Core is that it's an easy way to get yourself a fully primed AI box which requires no effort or setup.
This is for the AI enthusiast who doesn't want to buy a Mac mini (or Mac Studio) with maxed-out memory (required for larger AI models), and then have to mess around setting up everything themselves.
Javaid stresses that the PC comes with software built specifically for running those pre-installed AIs, avoiding all the fuss of model configuration. The CEO observes: "You can run [AI] models on existing hardware, but it's a terrible experience on many levels."
Core updates its models automatically as new upgrades emerge (unless you don't want that to happen), and the idea is that it's a seamless system and "similar to how Tesla works" with over-the-air updates.
You interact with the PC using an app (on your phone or laptop) with which you can issue voice commands should you wish.
Privacy and security
As well as the ease of use, privacy is the other big selling point for Core. It's tapping into a market of the increasing number of people realizing the potential of AI for helping with their life or work, who are concerned about sending their thoughts or data into the cloud, when they have no real idea where those queries or uploads will end up.
Core will run your AI agents continuously taking on whatever tasks you require while keeping all your data on the device, with Ghost promising that it's not going anywhere else. The personal data on the PC is encrypted to keep it safe — even Ghost can't access the data — and the device is specifically designed to keep running even if Ghost were to shut down (there's no subscription fee here).
The CEO assures us: "All of the source code, logic, and model weights live on the device, not the services. We build this device to not be dependent on having an outside entity maintaining the service."
Security is further bolstered by a firewall that limits how any agent can use credentials such as passwords, and you can dictate how much autonomy the Core agents are allowed, heavily restricting the AI should you prefer.
Javaid tells us that the firewall monitors all outgoing requests and "if it sees the agent performing a sanctioned action or exfiltrating data it shouldn't, it rejects the request and flags it for user review".
Core is priced at $3,499 in the US (around £2,600 / AU$5,000), and pre-orders for the first batch of PCs sold out on the first day that they were offered (yesterday). Those devices ship at the end of this month, and you can now register your interest in a second batch of PCs which are coming off the production line.
Analysis: a question of worth
Core can be hooked into all aspects of your life — or that's certainly the idea, and part of the reason why there's such an emphasis on privacy. It needs to be watertight for security given that Core connects to your existing devices and services, smart-home gadgets, fitness watches — or personal health data — and learns over time about how you live your life, imparting advice based on that knowledge.
Exactly how useful the system will be, well, that remains to be seen — and it may also depend on how 'smart' you've gone in terms of your array of gadgets and tech. There could be a lot of scope beyond life monitoring for Core's AI agents, of course, and there will presumably be the flexibility to get them doing whatever you want.
The other potential sticking point is that price. Is $3,499 a lot to pay for a dedicated AI PC? Well, it's undeniably a hefty sum of money, but there are some heavyweight components under the hood here (and the RAM crisis remains in full force). Indeed, the current price of the Nvidia GPU actually makes the PC a bargain as it stands (the company bought its stock of those graphics cards when they were much cheaper — so the next batch will be pricier).
If you configure a Mac mini as a top-of-the-line device for AI, you're looking at $3,199 in the US at the time of writing — and it doesn't come with the aforementioned strengths of software specifically designed for running local AI easily, and the privacy and security elements mentioned.
I've a feeling Ghost might be onto something here, at least in terms of pushing the concept of a user-friendly and straightforward AI box that does everything locally. I doubt this will be the last take on this kind of idea, and the price tag involved may eventually become a bit more realistic in the future. Although that depends on how the RAM crisis shakes out — and the latest signs are far from positive, sadly.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.
➡️ Read our full guide to the best laptops
1. Best overall:
Apple MacBook Air 13-inch M5
2. Best budget:
Apple MacBook Neo
3. Best Windows 11 laptop
Microsoft Surface Laptop 13-inch
4. Best thin and light:
Lenovo Yoga Slim 9i
5. Best Ultrabook
Asus Zenbook S 16