Ghost Raises $11M for Core Personal AI Computer

Personal AI becomes useful only after it learns what a person is doing, remembers what mattered, and earns permission to act. Ghost has raised an $11M Seed round to build that kind of agent around dedicated hardware rather than another account living entirely in someone else's cloud.

Andreessen Horowitz led the financing, with Abstract, Audacious Ventures, SV Angel, and Nova participating. The San Francisco startup emerged from stealth on October 5, 2026 with Core, a $3,499 screenless personal AI computer that Ghost says runs models, memory, and agent software locally.

The round puts capital behind a demanding thesis: the most capable personal agent will need broad access to a user's files, apps, wearables, cameras, and devices, but that access becomes harder to trust when the intelligence lives on remote infrastructure. Ghost is betting that ownership of the machine can become part of the product's value.

What Ghost Announced

Zain Javaid, Ghost's 19-year-old founder and CEO, announced the $11M raise alongside the launch of Core. TechCrunch reported that Ghost had spent roughly a year building before emerging from stealth and identified Nicholas Chua, Yifei Chen, and Gautam Sharda as the rest of the founding team.

Core is a full computer without an integrated screen. Users access it through a phone or app, while the device runs models and agent tasks in the background. Ghost's current product page lists an NVIDIA RTX PRO 4000 Blackwell SFF Edition GPU with 24 GB of memory, an AMD Ryzen 5 7600 CPU, 64 GB of system memory, and 1 TB of NVMe storage.

Ghost sells the first batch for $3,499 with no subscription. TechCrunch reported that shipments were scheduled for the last week of October, while Ghost's product page says the first batch is open for orders. No unit count, revenue figure, or independent demand metric has been disclosed.

Why Core Is a Hardware Bet

Core is less interesting as a specification sheet than as an argument about where personal intelligence should live. Cloud assistants can use powerful models and update quickly, but long-term personal context introduces a different set of expectations around privacy, control, availability, and ownership.

Ghost says Core connects to apps, files, screen history, wearables, cameras, and smart-home devices while keeping models and personal memory on the device. It is designed to observe patterns continuously and take action without waiting for a fresh prompt. The company also says the device can keep operating if Ghost shuts down because its code, logic, and model weights live locally.

Those claims turn local compute into a commercial promise. A buyer is not paying only for a GPU in a compact enclosure. The buyer is paying for a system that is supposed to combine memory, model management, integrations, a browser, a file system, and agent behavior without sending the most sensitive context back to a provider by default.

That is also where the product has to earn its premium. General-purpose computers can already run local models, and technically inclined users can assemble agent stacks on their own. Ghost must make the complete experience safer, easier, and more dependable than a collection of parts and open-source tools.

Trust Is the Product Constraint

Keeping inference and memory local can reduce exposure to cloud providers, but local architecture does not remove every security problem. An agent that reads email, connects to financial accounts, watches a camera feed, or orders something online still uses networks, credentials, and third-party services.

TechCrunch reported that Ghost was building a firewall to monitor outgoing requests and stop an agent from using credentials in unauthorized ways. That detail matters because it separates the product vision from the current proof. Ghost has described the control it wants, but the public record does not yet include an independent security audit, reliability benchmark, or outside evaluation of the agent's autonomy controls.

The same evidence boundary applies to performance. Ghost's current product page lists four preinstalled model families, although TechCrunch's launch report listed three. The exact catalog can change through software updates, so the stronger question is whether Core can select, run, and coordinate models well enough to make continuous local assistance feel useful rather than merely private.

What the $11M Needs to Prove

The investor group gives Ghost room to work across hardware supply, software, security, and the consumer experience, although the company has not disclosed a detailed use-of-funds allocation. Andreessen Horowitz has listed Ghost among its investments, and Audacious Ventures identifies both Ghost and Javaid in its portfolio.

That backing reflects a broader shift in consumer AI. The category is moving from occasional chat sessions toward agents that remember context, connect to services, and perform work. When an assistant becomes persistent, the placement of memory and permissions can matter as much as the underlying model.

Ghost now has to show that dedicated ownership changes the economics and behavior of personal AI. A $3,499 purchase with no subscription may appeal to buyers who value local control, but hardware brings manufacturing, support, upgrade, and reliability obligations that pure software companies can avoid.

The first real test will arrive in homes, not in the launch announcement. Core has to ship, stay useful after the first week, handle integrations without becoming fragile, and protect credentials while acting on a person's behalf. If Ghost can do that, the company will have made local ownership a meaningful part of the personal-AI market rather than a specification listed beside the GPU.

The $11M round finances that transition from premise to product. Ghost has chosen a difficult place to compete, where intimate context creates both the value and the risk, and where every useful action becomes another reason the machine must deserve the access it has been given.