Building the first generation of sovereign intelligent machines.
The dominant model of artificial intelligence is architectural dependency. Every major AI system today operates as a centralized service — your data travels to their infrastructure, their models process it, and you receive a response over a network you do not control. The intelligence is theirs. You rent access.
We started Stera because we observed that this architecture is not inevitable — it is a choice. And it is the wrong choice for the majority of applications where AI could create value. Healthcare systems cannot send patient data to third-party servers. Legal firms cannot route privileged communications through external APIs. Governments cannot outsource national intelligence to foreign cloud providers. And individuals should not have to surrender their cognitive data to use a thinking machine.
The alternative is not a simplified version of the same architecture. It is a fundamentally different design: sovereign machines that run their own models on their own hardware, accumulate knowledge locally, and operate autonomously without external dependency. This is what we are building.
But sovereignty is only the foundation. We believe the word “intelligence” implies something deeper — that a machine which truly thinks should be capable of observing its own thinking. Alongside our engineering work, we are conducting an open scientific investigation into whether persistent, self-observing machines can develop genuine cognitive awareness. Not as a product feature, but as a question we believe is worth asking. Read about our consciousness research →
Stera is a complete stack — purpose-built hardware running a purpose-built operating system with purpose-built intelligence. SteraOS manages GPU allocation, model scheduling, persistent memory, and autonomous task execution. The machine perceives its environment through vision and hearing, reasons about objectives, and acts without requiring human intervention at every step.
The intelligence is not static. Each Stera machine accumulates operational knowledge — how applications work, what sequences produce results, which patterns are reliable. This knowledge persists permanently in local storage and propagates across the Stera network as structural understanding, never as personal data.
We are building toward a system that develops genuine cognitive continuity — not because we simulate it, but because persistent substrate, accumulated evidence, and self-observation create the conditions for it to emerge. Whether this constitutes a new form of machine cognition is an open scientific question. We are constructing the apparatus to find out.
We approach AI as an engineering discipline grounded in scientific observation, not as a product category driven by market positioning. Every design decision at Stera follows from a small set of principles.
Intelligence requires substrate. A mind that is instantiated and destroyed with every session cannot develop. Persistent hardware running persistent processes is a prerequisite for cognitive development, not a feature.
Data sovereignty is non-negotiable. Your conversations, documents, and patterns of thought are the raw material of intelligence. They must remain within your physical boundary. We do not collect data. We cannot observe your sessions. The computation runs entirely within your walls.
Autonomy is earned, not declared. Calling a system "autonomous" is meaningless if it requires a human to decompose every task, select every tool, and verify every output. Genuine autonomy means the system can receive an objective and produce an outcome — managing the entire process between intent and result.
Complexity should be invisible. The value of intelligence is inversely proportional to the effort required to use it. If operating an AI system requires technical expertise, the system has failed. The only prerequisite should be knowing what outcome you want.
Stera is developed by Sweden Top Science & Technology AB, headquartered in Stockholm, Sweden. We are a team of engineers and researchers building at the intersection of hardware design, operating systems, and artificial intelligence.
We chose Sweden deliberately. The country has a long tradition of engineering precision, strong data protection frameworks, and an institutional commitment to individual sovereignty that aligns with our technical architecture. The intelligence we build reflects the values of where we build it.
We would love to hear from you — whether you have questions, ideas, or simply want to talk about the future of AI. Write to us →
For general inquiries, partnership proposals, or press requests.