Ghost Builds a $3,499 Dedicated Hardware Appliance for Personal AI Agents
Nineteen-year-old founder Zain Javaid has raised an $11 million seed round to launch Ghost, a startup building a $3,499 dedicated hardware appliance for personal AI agents.
According to TechCrunch, nineteen-year-old founder Zain Javaid has launched a startup named Ghost after a year of building, backed by an $11 million seed round led by Andreessen Horowitz. Additional investors in the funding round include Abstract, Audacious Ventures, SV Angel, and Nova. Javaid's founding team includes Nicholas Chua, 19, Yifei Chen, 24, and Gautam Sharda, 23. The company's premier offering is a specialized hardware appliance called Core, priced at $3,499, which is designed from the ground up to host personal AI agents capable of acting on a user's behalf.
Reporting by TechCrunch notes that Core functions as a screenless, desktop computing box that users access through a phone application, which also supports voice interaction through the app interface. The device is built to centralize a user's personal data from desktop files, applications, and smart home devices into a single localized unit. Javaid explained to TechCrunch that running models on existing hardware provides a poor user experience, prompting the team to create a dedicated physical architecture equipped with its own software layer, browser, and filing system.
According to TechCrunch, Core includes an Nvidia RTX Pro 4000 SFF Blackwell GPU bundled directly into the device's purchase price. The appliance comes with three models preinstalled, specifically Qwen-3.8-Next, Qwen-3.8-27B, and Gemma-4-31B. Furthermore, users have the flexibility to install alternative models from Hugging Face or load their own trained models onto the system. Javaid stated that the hardware and software are custom-built for running AI agents, distinguishing it from general-purpose small desktop computers like Apple's Mac Mini that require manual configuration for AI tasks.
TechCrunch reports that Core is designed to continuously process information and background tasks without requiring constant user prompts. Javaid provided an example where the device proactively notifies a user about potential illness by correlating an elevated resting heart rate, late-night computer work history, and seasonal flu context in San Francisco. The system receives improvements via over-the-air updates similar to a Tesla, allowing users to either automate updates or select when to transition to newer model generations as they become available.
According to TechCrunch, Ghost builds the units directly at a facility in San Francisco, with preorders opening on a Monday and shipments scheduled for the final week of October. Privacy and data ownership form core tenets of the hardware design, with local execution of personal memory and models. Javaid told TechCrunch that all personal information is encrypted so that third parties and Ghost cannot access it, and the device functions independently of outside maintenance services even if the company shuts down.
Reporting by TechCrunch indicates that Ghost is also developing a specialized firewall to oversee credential usage, such as logins and passwords. The firewall inspects every outgoing network request from the computer, automatically rejecting and flagging actions if it detects the agent performing an unauthorized task or attempting data exfiltration. Javaid emphasized that users retain complete ownership of the hardware, memory, and stored data, ensuring that everything remains processed locally on the device.