The sovereignty conversation is stuck
Data sovereignty is typically framed as a regulatory concern. Where is the data stored? Which jurisdiction governs it? Does it cross borders? These questions matter. But they miss the more fundamental issue. Sovereignty is not primarily about geography. It is about who controls the intelligence that data produces.
A European telco that sends its customer data to a US cloud provider for processing has technically lost sovereignty over the intelligence derived from that data. The data might be stored in Frankfurt. But the models that process it, the inferences they produce, and the value they create are controlled by the platform. The telco is renting insight from its own customers.
Intelligence is the asset. Not data.
Raw data depreciates. A behavioural signal from last week is less valuable than one from this morning. A customer interaction from six months ago tells you what someone wanted then, not what they want now. The data itself is a wasting asset.
Intelligence is different. The ability to interpret signals in real time, to detect patterns, to predict intent, and to act on that prediction before a competitor does. That is a durable competitive advantage. Companies that own their intelligence infrastructure own the future. Companies that outsource it are building on rented ground.
On-device processing changes the equation
When intelligence is processed on the device, several things happen simultaneously. The raw data never moves, so there is no jurisdiction question. The intelligence is produced locally, so there is no dependency on a third-party cloud. And the company deploying the on-device model retains full control over the algorithms, the outputs, and the strategic value they produce.
Intent builds on-device AI that runs on the handset. The models process behavioural signals locally and produce privacy twins that carry intelligence without identity. The brand or telco deploying Intent owns the intelligence. It does not share it with a cloud provider. It does not expose it to foreign jurisdiction. It does not create a dependency that could be severed by a policy change or a geopolitical shift.
The competitive case
Consider two telcos competing in the same market. Telco A sends its data to a hyperscaler for processing. Telco B processes on-device with Intent. Both have access to the same customer base.
Telco A gets its intelligence back on a 24-hour cycle. The models are shared infrastructure, trained on aggregated data from multiple clients. The insights are useful but generic. The telco cannot customise the models without paying for dedicated compute. And the hyperscaler has visibility into the telco’s data patterns, which it can use to build competing products.
Telco B gets intelligence in real time. The on-device models are purpose-built for its market. The behavioural signals are fresher. The privacy architecture means regulatory exposure is minimal. And no third party has visibility into the telco’s intelligence capabilities or the signals that drive them.
Telco B has a structural advantage that Telco A cannot close by spending more on cloud compute. The advantage is architectural.
Sovereignty as strategy
The companies that will lead their categories over the next decade are those that own their intelligence. Not their data. Their intelligence. The ability to interpret, predict, and act on behavioural signals in real time, without dependency on a platform that could reprice, restrict, or replicate that capability.
Sovereign AI is not a compliance checkbox. It is not a moral position. It is a competitive architecture. The companies that build it will set the pace. The companies that outsource it will follow.
The geopolitical dimension
This is not abstract. The US CLOUD Act gives American authorities the ability to compel US-headquartered cloud providers to produce data stored anywhere in the world. The EU Data Act and AI Act are creating new obligations around data access and algorithmic transparency. China’s data localisation requirements continue to tighten.
Companies that process intelligence on-device are insulated from all of these pressures. The data never enters a cloud. The intelligence is produced locally. There is nothing to compel, nothing to localise, nothing to audit. The architecture resolves the regulatory question before it arises.
Sovereignty is increasingly a prerequisite for operating at scale in regulated industries. The companies that treat it as a competitive weapon rather than a compliance burden will find they have built something their competitors cannot easily replicate.