Published: October 6, 2026 · Last updated: October 6, 2026
TL;DR: Ghost, a startup founded by 19-year-old Zain Javaid, came out of stealth today with an $11 million seed round led by Andreessen Horowitz to sell Core, a $3,499 computer built to run AI models on your desk instead of in someone else’s data center.
Most people spend their last year as a teenager worrying about college applications. Zain Javaid spent it building a computer.
His company, Ghost, announced today that it closed an $11 million seed round led by Andreessen Horowitz to build Core, a dedicated personal computer whose entire reason for existing is running AI models locally instead of shipping every prompt off to a data center somewhere else.
The pitch makes sense once you hear it. Every mainstream AI assistant, ChatGPT, Gemini, Claude, runs its heavy lifting on someone else’s servers. You type something, it travels to a data center, a model chews on it, and an answer comes back a second or two later. That round trip costs the company money every single time, raises obvious privacy questions, and depends entirely on your internet connection and whoever’s servers are on the other end staying up. Core is built to skip most of that. The model lives on your desk. The hardware is sized specifically so it doesn’t choke trying to run it.
That focus comes at a real cost, and not just the $3,499 price tag. Ghost itself says Core still needs an internet connection for things like live web search, since no model sitting in a box under your monitor can index the entire internet in real time. What it doesn’t need the internet for is the thinking itself. Ask it something that doesn’t require up-to-the-minute information and the answer comes from silicon a few feet away, not a server farm in Virginia.
It’s a bet that a meaningful slice of AI users, developers, privacy-conscious professionals, people in regulated industries who legally can’t pipe client data through a third-party API, would rather own their intelligence outright than rent it by the token. That’s a real market. It’s also one that Apple, Nvidia, and a growing list of smaller hardware startups are circling at the same time, all chasing some version of “AI that doesn’t phone home.” A teenager with an a16z check just put a very specific price on that idea, and the market now gets to decide if $3,499 is a bargain or a niche product looking for a niche that’s smaller than the pitch deck promised.
Ghost says Core ships with its own operating system built specifically for running open-weight models like Llama and DeepSeek without a line of configuration from the buyer, which matters more than it sounds. Most people who want to run a large language model on their own hardware right now have to piece together drivers, a runtime, and a model file themselves, and most people never bother. Javaid’s bet is that the hardware and the software need to ship as one product, priced once, rather than as a build-it-yourself kit for people who already know what a GPU driver is.
Andreessen Horowitz has been vocal for more than a year about backing what it calls the local AI category, arguing that running models on a device will eventually matter to consumers the way running software locally mattered before cloud computing took over. Ghost is one of several bets the firm has made on that thesis, alongside stakes in AI-focused chips and edge-inference startups. Whether that thesis holds depends on model sizes staying small enough to run well on consumer-grade hardware, which isn’t guaranteed as the industry keeps pushing toward bigger, more capable systems.
Related: Flai just raised $27 million automating car dealership phone calls and Amazon’s 2026 Kindle lineup brings physical buttons back.
Bottom Line: Ghost is betting that owning your AI outright beats renting it from a cloud provider, and a16z just backed that bet with real money. Whether $3,499 buyers show up in numbers large enough to matter is the only question here that actually counts.
Follow Teck Hustlers: Facebook · X · Instagram · LinkedIn · TikTok · Pinterest
