Olares, a new entrant in the consumer AI hardware space, has announced Olares One , a personal cloud server and AI workstation designed to keep data private while training a personal assistant on local files. The pitch behind Olares One reflects a growing tension in consumer technology . AI apps are becoming indispensable, from productivity platforms to creative tools, yet they increasingly rely on access to sensitive data. Olares argues that the current model asks too much. Personal documents and behaviors help train corporate models, while users receive only a fraction of the value in return. The company’s answer is to bring AI processing back to the device itself, offering performance and automation without necessarily sending your data to the cloud.
Turning your files into a private knowledge base
Olares One works by organizing documents, messages, and notes into a local knowledge base. Rather than juggling apps and services, users create what could be considered a digital twin of their work and personal lives. The system uses this content to train a local assistant that learns writing style, workflow patterns, and personal preferences.
In theory, this turns everyday data into a context-aware engine that understands the user as an individual because it has access to their information, not anonymized behavior from millions of people. The result is a personalized assistant that lives entirely on-device.
Designed to work like a normal PC, but built for Speed
Privacy-centered hardware often comes with technical barriers. Olares says its device is different. Olares One uses an open source operating system but is designed to run like a conventional PC. Users can download and install applications with a simple click, and the system handles security behind the scenes.
A Remote Desktop Client lets users connect their current Windows PC or laptop and run their software through Olares One as though installed locally. This matters because it lowers the barrier for people who want better performance without switching their workflow.
Security is built into the architecture rather than bolted on. App Sandboxing isolates software so that apps cannot access unrelated data without explicit permission. Even if an app misbehaves, the damage is contained. It is privacy by default, rather than privacy that depends on user settings.
While the privacy pitch is attention-grabbing, Olares spends just as much time highlighting performance. Olares One is fitted with an NVIDIA GeForce RTX 5090 Mobile GPU with 24GB of high-bandwidth memory, paired with an Intel Core Ultra 9 275HX processor and 96GB of RAM.
The company claims this setup can outperform other personal AI systems in LMSys benchmarks across models like GPT, Llama, Qwen, Deepseek, and Gemma. The goal is to run several AI tools at once directly on the device, without offloading workloads to remote servers.
A notable capability is Time Sharing Mode, which dynamically allocates GPU resources between applications. Instead of locking resources to a single program, Olares One tries to maximize efficiency for multitasking and heavy workloads.
Users can even run advanced AI tools from a browser while keeping all data stored locally. It is positioned as both a private server and an AI powerhouse for people who work with video, images, or large datasets.
The big question is whether consumers will embrace local computing in an era defined by cloud services. With Olares One launching on Kickstarter today for US $2999, Olares believes the answer is yes, and Olares One is its opening move in a category that is still waiting for its watershed moment.
