Startup Ghost Secures $11M to Build Dedicated Local Hardware for AI Agents
AI hardware startup Ghost AI has raised $11 million led by Andreessen Horowitz to develop Core, a dedicated home computer built exclusively to run private AI agents locally.
Artificial intelligence hardware startup Ghost AI announced that it has successfully raised $11 million in a funding round led by Silicon Valley firm Andreessen Horowitz. Additional participants in the round included Abstract, Audacious Ventures, Nova, and SV Angel. Ghost AI is led by a team of young entrepreneurs, including 19-year-old Chief Executive Zain Javaid. Over the past year, the company has focused on developing a powerful computer named Core, designed specifically to operate personal AI agents without requiring a monitor or other peripheral equipment.
The Concept Behind Core
Described as a "brain in a box," Core is engineered to continuously run AI agents on behalf of its user. The founders at Ghost argue that existing cloud-based AI agent infrastructure, typically managed by large technology corporations, contains fundamental flaws. Because effective AI agents require access to massive quantities of deeply personal and private user data, reliance on third-party cloud infrastructure demands a high level of trust. In an initial post on X, the startup emphasized that an intelligence system with an intimate understanding of a user's life should not reside on external infrastructure.
To address these privacy concerns, Javaid and his co-founders designed a locally hosted machine that users can install directly in their homes. By keeping operations local, AI agents can continuously learn, analyze private data, and execute tasks without transmitting data fragments to the cloud.
Hardware and Software Specifications
Core features an Nvidia RTX Pro 4000 SFF Blackwell graphics processing unit to supply agents with substantial processing power. The device includes its own proprietary software layer, file system, and web browser. At launch, it ships with multiple open-source models, such as Alibaba Group Holding's Qwen-3.8, Qwen-3.8-2.7B, and Google LLC's Gemma-4-31B. Users also retain the option to download alternative models from Hugging Face and other repositories.
Because the device lacks a screen or conventional consumer operating system like Windows or Ubuntu, users interact with it via a dedicated smartphone and web application. This control center allows users to message or speak to their personal agents to assign tasks and issue commands. Over-the-air updates are delivered by Ghost when the device connects to the web.
Security remains a primary focus for the hardware. With all models and user data contained within the device, storage is fully encrypted, and users maintain the encryption keys so that no third party, including Ghost, can access the stored contents. Furthermore, a built-in firewall monitors outgoing agent requests to prevent data exfiltration. The system is also designed for longevity; because the source code, logic, and model weights run directly on the machine, the computer would continue to function even if Ghost ceased operations.
What it means for developers
For developers exploring local agent workflows, hardware like Core points toward a growing demand for decentralized, privacy-focused AI deployment where weights and logic stay on-device. For those working across cloud environments, developers can try top AI models cheaply through one API at https://apixoai.online.
Despite a price tag of $3,499 per unit, the first batch of Core devices sold out within hours. With $11 million in fresh funding, Ghost is now inviting users to register interest for its upcoming second production batch.
Source: AI agent hardware startup Ghost, led by its 19-year-old-founder, raises $11M — SiliconANGLE AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
