Ghost raises $11 million for Core, a $3,499 personal AI computer
The AMW Read
Ghost introduces a dedicated local-hardware approach to personal agents, meaningfully expanding the player map, but its early launch and unvalidated claims limit demonstrated impact to a sub-segment.
Ghost raises $11 million for Core, a $3,499 personal AI computer
Ghost emerged from stealth on October 5 with an $11 million seed round led by Andreessen Horowitz, alongside Abstract, Audacious Ventures, SV Angel, and Nova. Its first product, Core, is a screenless computer designed to continuously run personal AI agents, accessed through a phone or app. The $3,499 device includes an Nvidia RTX Pro 4000 SFF Blackwell GPU and three preinstalled models: Qwen-3.8-Next, Qwen-3.8-27B, and Gemma-4-31B. Preorders opened that day, with shipping scheduled for the final week of October. The team builds units at its San Francisco facility.
Core adds a dedicated-device approach to the personal-agent market, bundling local inference with software for memory, browsing, files, and actions. The competitive question is whether that integration makes agents sufficiently useful and simple to justify buying separate hardware. CEO Zain Javaid says models and personal memory run locally, personal information is encrypted, and the device can keep working if Ghost shuts down. Those are company claims; the supplied report provides no independent performance or security validation. Local storage also leaves a separate challenge: controlling what agents can do through connected accounts.
For investors, the concrete test is whether Ghost can turn a hardware purchase into sustained agent usage. Evaluation should examine setup time, task completion, and how reliably users can limit autonomy. Javaid says the team is building a firewall to monitor outgoing requests and restrict agents' use of credentials, making its effectiveness an important diligence item. Builders face a similar hurdle: preinstalled models simplify deployment, but differentiation depends on dependable memory and authorized actions across a user's applications, rather than the GPU alone.